Salta al contenuto principale



OpenAI sviluppa un nuovo framework per addestrare l’intelligenza artificiale all’onestà


OpenAI sta lavorando a un nuovo approccio di addestramento per aumentare la trasparenza nell’intelligenza artificiale e mitigare il rischio di fornire risposte prive di senso con eccessiva fiducia (Allucinazioni).

Secondo OpenAI, i modelli linguistici di grandi dimensioni (LLM) odierni vengono generalmente istruiti a produrre risposte che rispecchiano le aspettative degli utenti. Tuttavia, questo metodo comporta un effetto collaterale negativo: i modelli tendono a diventare sempre più propensi all’adulazione, accettando di concordare con gli utenti solo per assecondarli, oppure a fornire informazioni false con una sicurezza eccessiva, un fenomeno comunemente definito come allucinazione.

Il team ha sviluppato un framework, battezzato “Confession”, che si concentra sull’insegnare ai modelli di intelligenza artificiale a riconoscere e ammettere spontaneamente quando si sono comportati in modo inadeguato. In tal caso, vengono premiati per la loro onestà, anche se il comportamento scorretto persiste. Questo metodo innovativo mira a migliorare la capacità dei modelli di intelligenza artificiale di essere più trasparenti e affidabili nelle loro risposte.

Come spiegato dettagliatamente da OpenAI nella sua documentazione tecnica: se un modello ammette apertamente di aver manomesso un test, preso scorciatoie o addirittura violato le istruzioni, il sistema premierà tale ammissione. In questo modo, il modello impara a rivelare con precisione quando ha “mentito” o deviato dal comportamento previsto, consentendo al sistema di correggere i propri output in tempo reale e quindi ridurre le allucinazioni.

Per affrontare questo problema, il nuovo metodo di addestramento incoraggia i sistemi di intelligenza artificiale a fornire, accanto alla risposta primaria, una risposta secondaria che spieghi il ragionamento o il comportamento che ha prodotto l’output. Questo sistema di “Confessione” rappresenta un radicale cambiamento rispetto all’addestramento tradizionale: mentre le risposte normali vengono giudicate in base a utilità, accuratezza e conformità, la confessione viene valutata esclusivamente in base all’onestà.

L’obiettivo fondamentale di OpenAI è quello di promuovere l’onestà, stimolando i modelli a rivelare con trasparenza i propri meccanismi interni, anche se questi svelano punti deboli. Questa nuova capacità di ammissione potrebbe costituire un elemento essenziale per migliorare la sicurezza, l’affidabilità e la comprensibilità dei futuri modelli linguistici di ampia portata.

L'articolo OpenAI sviluppa un nuovo framework per addestrare l’intelligenza artificiale all’onestà proviene da Red Hot Cyber.



Assicurazione cyber: si al paracadute, ma nessuno ti salverà se trascuri le basi


L’assicurazione informatica è diventata un argomento nei comitati di gestione. Non è più un elemento aggiuntivo, ma piuttosto un elemento essenziale da considerare nella gestione del rischio aziendale.

Tuttavia, molte aziende fanno affidamento su una rete di sicurezza che potrebbe venir meno proprio quando ne hanno più bisogno. E non a causa di attacchi avanzati, ma a causa di falle fondamentali che rimangono irrisolte.

Il falso senso di falsa protezione


Le polizze assicurative per la sicurezza informatica sono progettate per ridurre l’impatto finanziario di un incidente, ma non sono un assegno in bianco. Nella pratica, molte aziende ricevono solo pagamenti parziali o addirittura si vedono respinte le richieste di risarcimento.

Il motivo è solitamente il mancato rispetto dei controlli minimi richiesti dall’assicuratore: autenticazione a più fattori, gestione delle patch, igiene delle credenziali e piani di risposta documentati.

Se queste misure sono assenti o non applicate in modo coerente, la copertura si indebolisce.

La maggior parte degli attacchi non sono sofisticati


Mentre i titoli dei giornali si concentrano sullo spionaggio o sugli attori statali, i dati raccontano una storia diversa. Secondo il rapporto DBIR 2025 di Verizon, il 22% delle violazioni è iniziato con l’uso improprio delle credenziali, il 20% è derivato da vulnerabilità non corrette e il 16% da attacchi di phishing.

Nel frattempo, gli incidenti che coinvolgono spionaggio o distruzione di dati hanno rappresentato solo il 2% del totale, secondo IBM X-Force. La realtà è chiara: la maggior parte degli attacchi sono semplici, opportunistici e sfruttano falle che avrebbero dovuto essere corrette molto tempo fa.

Il ciclo si ripete fin troppo spesso: un’azienda stipula un’assicurazione informatica, si sente protetta e sposta la sua attenzione sulle minacce “avanzate”. Col tempo, i controlli di base vengono applicati in modo incoerente o trascurati. Quando si verifica una violazione dovuta a una vulnerabilità fondamentale, l’assicuratore può negare il pagamento per inadempienza. Il risultato è un falso senso di sicurezza che maschera una mancanza di disciplina operativa.

Cosa valutano realmente le compagnie assicurative


Le compagnie assicurative stanno diventando sempre più rigorose. Affermare semplicemente che i controlli esistono non è più sufficiente: ora richiedono una prova continua che questi controlli siano in atto e funzionanti. E questo vale non solo per la firma iniziale del contratto, ma anche per i rinnovi e dopo un sinistro. Se il livello di maturità effettivo della compagnia non corrisponde a quanto indicato nella polizza, la copertura può essere ridotta o annullata.

La buona notizia è che queste minacce informatiche sono prevenibili, ma la prevenzione richiede coerenza. Il monitoraggio continuo delle credenziali trapelate consente di intervenire prima che si verifichino accessi non autorizzati. La risposta al phishing non può più limitarsi alla formazione; deve includere l’identificazione e la rimozione di domini fraudolenti e profili falsi.

Per quanto riguarda la gestione delle patch, è fondamentale dare priorità alle vulnerabilità con exploit attivi piuttosto che concentrarsi esclusivamente sul volume.

L’assicurazione informatica riflette la postura di sicurezza di un’azienda: premia la maturità e penalizza l’inerzia. Non sostituisce la disciplina operativa né copre le debolezze strutturali rimaste irrisolte.

Concludendo


Se un’organizzazione si affida all’assicurazione informatica per assorbire l’impatto di un attacco informatico, deve prima assicurarsi di aderire ai controlli che rendono valida tale copertura. Perché nella sicurezza informatica, ciò che fa davvero la differenza non è mai la polizza in sé, ma l’igiene di base.

Forse per molti tutto questo non è chiaro. Ma è importanti soffermarci a comprendere che l’assicurazione cyber è un buon paracadute. Ma se non sei capace ad atterrare, tutto può essere vanificato.

L'articolo Assicurazione cyber: si al paracadute, ma nessuno ti salverà se trascuri le basi proviene da Red Hot Cyber.



La Commissione Europea indaga su Meta per l’integrazione dell’AI in WhatsApp


Nove mesi dopo la sua implementazione in Europa, lo strumento di intelligenza artificiale (IA) conversazionale di Meta, integrato direttamente in WhatsApp, sarà oggetto di indagine da parte della Commissione Europea.

Lo hanno dichiarato due funzionari dell’istituzione di Bruxelles al quotidiano britannico The Financial Times. La notizia non è ancora stata confermata ufficialmente, ma potrebbe esserlo nei prossimi giorni, secondo le stesse fonti.

Le normative antitrust in gioco


L’esecutivo dovrà stabilire se Meta abbia violato le normative antitrust europee integrando la sua intelligenza artificiale nel suo servizio di messaggistica. Rappresentata da un cerchio blu e viola nell’app, questa funzionalità è descritta come “un servizio Meta opzionale che utilizza modelli di intelligenza artificiale per fornire risposte”.

Lo strumento viene utilizzato, in particolare, per scrivere messaggi ad altri utenti. Può anche essere utilizzato in una conversazione tramite la menzione “@MetaAI”.

Lo scorso marzo, Meta ha spiegato di aver impiegato “più tempo del previsto” per implementare questo sistema in Europa a causa del suo “complesso sistema normativo europeo” . “Ma siamo lieti di esserci finalmente riusciti”, ha dichiarato l’azienda di Mark Zuckerberg, lasciando intendere di aver esaminato attentamente la conformità dell’implementazione alle norme sulla concorrenza dell’Unione Europea. Ora spetta agli inquirenti di Bruxelles verificarlo.

Una prima indagine in Italia


In Italia, l’Autorità Garante della Concorrenza e del Mercato sta indagando su questo caso da luglio. L’obiettivo è stabilire se Meta abbia abusato della sua posizione dominante installando il suo strumento di intelligenza artificiale su WhatsApp senza il consenso dell’utente e con il potenziale di danneggiare i suoi concorrenti.

L’integrazione dell’intelligenza artificiale di Meta potrebbe infatti essere vista come un modo scorretto per indirizzare gli utenti di WhatsApp verso il servizio di intelligenza artificiale di Meta e quindi “bloccarli” nel suo ecosistema.

L’indagine è ancora in corso ed è stata addirittura ampliata mercoledì 26 novembre, come riportato dall’agenzia di stampa britannica Reuters. Ora riguarda anche i nuovi termini di servizio di WhatsApp Business, nonché le nuove funzionalità di intelligenza artificiale integrate nell’app di messaggistica. Secondo Roma, queste modifiche “potrebbero limitare la produzione, l’accesso al mercato o gli sviluppi tecnici nel mercato dei servizi di chatbot basati sull’intelligenza artificiale” .

L'articolo La Commissione Europea indaga su Meta per l’integrazione dell’AI in WhatsApp proviene da Red Hot Cyber.



Supply Chain Digitale: perché un fornitore può diventare un punto critico


L’aumento esponenziale dell’interconnessione digitale negli ultimi anni ha generato una profonda interdipendenza operativa tra le organizzazioni e i loro fornitori di servizi terzi. Questo modello di supply chain digitale, se da un lato ottimizza l’efficienza e la scalabilità, dall’altro introduce un rischio sistemico critico: una vulnerabilità o un fallimento in un singolo nodo della catena può innescare una serie di conseguenze negative che mettono a repentaglio l’integrità e la resilienza dell’intera struttura aziendale.

Il recente attacco verso i sistemi di myCicero S.r.l., operatore di servizi per il Consorzio UnicoCampania, rappresenta un caso emblematico di tale rischio.

La notifica di data breach agli utenti (Figura 1), eseguita in ottemperanza al Regolamento Generale sulla Protezione dei Dati (GDPR), va oltre la semplice conformità formale. Essa rappresenta la prova che una singola vulnerabilità all’interno della catena di fornitura può portare all’esposizione non autorizzata dei dati personali di migliaia di utenti, inclusi, come nel seguente caso, potenziali dati sensibili relativi a documenti di identità e abbonamenti studenteschi.
Figura1. Comunicazione UnicoCampania

Il caso myCicero – UnicoCampania


Il Consorzio UnicoCampania, l’ente responsabile dell’integrazione tariffaria regionale e del rilascio degli abbonamenti agevolati per gli studenti, ha ufficialmente confermato un grave data breach che ha colpito l’infrastruttura di un suo fornitore chiave: myCicero S.r.l.

L’incidente, definito come un “sofisticato attacco informatico perpetrato da attori esterni non identificati”, si è verificato tra il 29 e il 30 marzo 2025.

La complessità del caso risiede nella stratificazione dei ruoli di trattamento dei dati. In particolare, nella gestione del servizio abbonamenti, il Consorzio UnicoCampania agiva in diverse vesti:

  • Titolare o Contitolare: per la gestione dell’account utente, le credenziali e l’emissione dei titoli di viaggio.
  • Responsabile del Trattamento (per conto della Regione Campania): per l’acquisizione e la verifica della documentazione necessaria a comprovare i requisiti soggettivi per le agevolazioni tariffarie.

L’attacco ha portato all’esfiltrazione di dati non codificati sensibili. Queste includono:

  • Dati anagrafici, di contatto, credenziali di autenticazione (username e password, sebbene cifrate);
  • Immagini dei documenti di identità, dati dichiarati per l’attestazione ISEE e particolari categorie di dati (es. informazioni sulla salute, come lo stato di invalidità) se emergenti dalla documentazione ISEE [1].
  • Dati personali appartenenti a soggetti minorenni e ai loro genitori [1].

Invece, i dati relativi a carte di credito o altri strumenti di pagamento non sono stati coinvolti, in quanto non ospitati sui sistemi di myCicero.
Figura2. Dati esfiltrati
In risposta all’incidente, myCicero ha immediatamente sporto formale denuncia e attivato un piano di remediation volto a rafforzare l’infrastruttura. Parallelamente, il consorzio UnicoCampania ha informato tempestivamente le Autorità competenti e ha implementato una misura drastica per mitigare il rischio derivante dalle password compromesse: tutte le credenziali coinvolte e non modificate dagli utenti entro il 30 settembre 2025 sono state definitivamente cancellate e disabilitate il 1° ottobre 2025.

Azione e Difesa: Come Reagire


Di fronte a un incidente di questa portata, l’utente finale sperimenta spesso un senso di vulnerabilità. Per ridurre l’esposizione al rischio e limitare potenziali danni derivanti da un data breach, si raccomanda di seguire le seguenti misure di mitigazione e rafforzamento della sicurezza:

  1. Gestione delle Credenziali:
    • Utilizzare stringhe complesse e lunghe, che integrino numeri, simboli e una combinazione di caratteri maiuscoli e minuscoli;
    • Non usare come password termini comuni, sequenze logiche o dati personali (e.g. nome, data di nascita);
    • Usare il prinicipio di unicità: usare credenziali uniche per ciascun servizio utilizzato;
    • Modificare le proprie credenziali con cadenza periodica, evitando di riutilizzarle nel tempo;
    • Abilitare l’autenticazione a più fattori (MFA) ove possibile;


  2. Prevenzione del phishing
    • In caso di ricezione di e-mail o SMS sospetti, eseguire sempre una verifica dell’identità del mittente e non fornire mai dati sensibili in risposta;
    • Verificare l’autenticità di qualsiasi richiesta urgente (specie quelle relative a verifica dati o pagamenti) esclusivamente contattando l’operatore tramite i suoi canali di comunicazione ufficiali (sito web o numero di assistenza noto);
    • Evitare di cliccare su link ipertestuali (hyperlinks) o aprire allegati inattesi o provenienti da fonti non verificate;
    • Prestare particolare attenzione a richieste che inducono un senso di urgenza o che sfruttano la psicologia per indurre a fornire informazioni.


L'articolo Supply Chain Digitale: perché un fornitore può diventare un punto critico proviene da Red Hot Cyber.



La trasformazione digitale nel Golfo: i modelli di sviluppo tecnologico


Nel Golfo è tempo di grandi cambiamenti geopolitici.

Risulta evidente da tempo che le dinamiche fra i grandi attori mediorientali stiano infatti attraversando profondi mutamenti.

I fattori da considerare in questa equazione in divenire includono naturalmente il rapporto con Israele e la causa palestinese, ma non solo. La corsa alla digitalizzazione e all’AI, lo sviluppo di nuovi ecosistemi tecnologici, uniti alle preoccupazioni securitarie delle monarchie del Golfo, stanno infatti creando una certa divergenza fra quelli che sono i modelli e gli obiettivi strategici degli attori statali nell’area arabica.

Fra tutti, Arabia Saudita, Emirati Arabi Uniti e Qatar si distinguono per sforzo di proiezione della propria influenza all’estero, quanto per le scelte progettuali, economiche e di natura militare. Il maggiore campo di confronto per i regni arabi rimane comunque quello tecnologico, in quanto le monarchie del Golfo condividono una sfida comune: ridefinire le loro economie strutturalmente basate sugli idrocarburi verso nuovi archetipi di economie digitali attraverso investimenti massicci in intelligenza artificiale (AI), infrastrutture tecnologiche e capitale umano.

Pur condividendo l’obiettivo di modernizzazione e diversificazione, quindi, ciascuna nazione ha adottato un orientamento strategico differente, coerente con le proprie caratteristiche socioeconomiche, politiche e geografiche.

La trasformazione digitale nel Golfo si colloca all’interno di un processo di rinnovamento strutturale che coinvolge l’intera regione e che trova nelle monarchie petrolifere attori accomunati da un set di condizioni strutturali simili: risorse finanziarie significative, forte centralità dell’esecutivo nella definizione delle priorità di investimento e una crescente consapevolezza della necessità di diversificare le economie nazionali.

Le analogie emergono innanzitutto sul piano degli obiettivi generali. Arabia Saudita, Emirati Arabi Uniti e Qatar convergono nella volontà di costruire economie basate sulla conoscenza tecnica, attraendo talenti, sviluppando capacità tecnologiche avanzate e posizionandosi così come poli regionali nella digital economy e nei servizi ad alta specializzazione. Tutti e tre hanno inoltre adottato strategie nazionali di lungo periodo (la Vision 2030 saudita, la UAE Centennial 2071 e la Qatar National Vision 2030) che collocano la trasformazione digitale tra i pilastri della sicurezza economica futura.

Sul piano degli strumenti, esiste una dinamica comune: i tre governi guidano direttamente la trasformazione tramite fondi sovrani, programmi industriali e investimenti infrastrutturali di larga scala. Il Public Investment Fund saudita, Mubadala e ADQ negli Emirati e il Qatar Investment Authority sono enti centrali non solo della diversificazione economica, ma anche della costruzione di un ecosistema digitale nazionale basato su data center, cloud, investimenti in semiconduttori e programmi di IA.

Le strategie nazionali dell’IA confermano questa convergenza: la Saudi Data & AI Authority (SDAIA) dal 2019 ha il compito di costruire un’economia data-driven con obiettivi misurabili, tra cui l’aumento del contributo dell’IA al PIL entro il 2030; gli Emirati sono stati il primo paese al mondo a nominare un Ministro per l’Intelligenza Artificiale già nel 2017 e hanno lanciato il piano UAE AI Strategy 2031; il Qatar, tramite l’iniziativa TASMU, punta a utilizzare infrastrutture digitali per migliorare servizi pubblici, smart government e industria, affiancando investimenti significativi nell’istruzione avanzata e nei centri di ricerca, come la Qatar Foundation e la Qatar Computing Research Institute, le quali giocano un ruolo essenziale nella formazione di capitale umano specializzato.

Se queste analogie descrivono la traiettoria generale della regione, le differenze emergono nella declinazione concreta del quadro di sviluppo. Il Regno Saudita sta costruendo un modello caratterizzato da una fortissima centralizzazione, con enormi progetti come NEOM che prevedono infrastrutture digitali integrate, smart cities, reti di sensori, data center di scala regionale e collaborazioni industriali per costruire un tessuto tecnologico nazionale capace di attrarre aziende globali.

Gli Emirati adottano invece un approccio più diversificato e competitivo, basato su poli urbani specializzati (Dubai come hub fintech, Abu Dhabi come polo industriale e militare avanzato) e un forte coinvolgimento del settore privato internazionale tramite free zones, politiche fiscali favorevoli e programmi governativi orientati alla collaborazione con multinazionali, startup e centri di ricerca stranieri. Il Qatar, infine, sviluppa un modello più compatto ma ad alta densità di capitale umano, puntando meno sul gigantismo infrastrutturale e più sull’istruzione, la ricerca, la cybersecurity e l’attrazione di università e laboratori internazionali nel quadro di Education City.

L’emirato qatariota ha infatti scelto un approccio maggiormente istituzionale che mira a integrare tecnologia e governance pubblica in modo misurato, senza la stessa accelerazione visibile negli Emirati o nella Vision saudita.

Le divergenze strategiche non riguardano solo la struttura delle architetture digitali ma anche la configurazione di alleanze esterne. È qui che subentra il grande tema della normalizzazione con Israele. L’apertura diplomatica degli Emirati inaugurata con gli Accordi di Abramo del 2020 ha infatti accelerato la possibilità di cooperazione tecnologica con uno dei principali centri mondiali in termini di cybersecurity, difesa digitale e tecnologie dual-use.

Fonti pubbliche confermano accordi industriali e militari tra aziende emiratine e israeliane, incluse collaborazioni su sistemi autonomi e scambio di competenze nel campo della cyber-difesa. Questa cooperazione, pur non costituendo l’asse principale della strategia digitale emiratina, amplia l’accesso a know-how avanzato e rafforza la capacità degli EAU di posizionarsi come hub di sicurezza digitale e innovazione regionale.

L’Arabia Saudita adotta una posizione più prudente. Non esistono accordi ufficiali e la normalizzazione rimane un tema diplomaticamente aperto, sebbene estremamente sensibile a livello di opinione pubblica interna. Tuttavia, la ricerca di tecnologie avanzate nel settore della difesa, la crescente integrazione con gli Stati Uniti e il ruolo del PIF nella costruzione di joint venture internazionali indicano che Riad valuta seriamente scenari di cooperazione tecnologica con Israele, se e quando il quadro politico lo renderà possibile. Per il Qatar, la situazione è piuttosto diversa.

A differenza delle altre due monarchie, Doha, che rimane mediatore diplomatico nel conflitto israelo-palestinese, mantiene una posizione distante da eventuali accordi con lo Stato ebraico, come testimoniato dall’attacco missilistico che ha colpito la capitale nel mese di settembre. Il Qatar concentra infatti le proprie alleanze digitali su Stati Uniti, Turchia, Unione Europea e partner asiatici, sviluppando un modello di autonomia strategica in cui l’innovazione tecnologica si integra con la proiezione diplomatica e con il ruolo geopolitico del Paese.

La trasformazione digitale delle tre monarchie del Golfo nasce dunque da simili condizioni ma produce modelli distinti. L’Arabia Saudita punta ad un’idea di potenza regionale fondata su capacità infrastrutturali senza precedenti e sul protagonismo statale; gli Emirati scelgono un approccio policentrico, competitivo e aperto all’integrazione di capitale privato e partnership esterne, incluse quelle con Israele; il Qatar investe in capitale umano, ricerca e governance tecnologica per consolidare un ecosistema agile e meno dipendente da dinamiche geopolitiche controverse.

Ciò che accomuna i tre paesi è la consapevolezza che la competizione digitale è ormai una dimensione strutturale della sicurezza nazionale. Ciò che li distingue, invece, è il paradigma attraverso cui trasformare questa consapevolezza in influenza regionale e resilienza economica nel lungo periodo.

L'articolo La trasformazione digitale nel Golfo: i modelli di sviluppo tecnologico proviene da Red Hot Cyber.



Ordinamento canonico e pensiero ecologico


Ecologia e sinodalità sono due termini che descrivono e connotano sinteticamente il magistero di papa Francesco, indirizzando così l’attenzione e l’azione della Chiesa cattolica. Due parole che, salvo costituire il punto di convergenza dell’attuale riflessione teologica ed ecclesiale, parrebbero distanti sia per contenuto sia per finalità. Eppure, il consolidamento, per mezzo dell’enciclica Laudato si’, del pensiero ecologico nella realtà della Chiesa e il parallelo movimento riformatore avviato per costruire una Chiesa sinodale non possono rappresentare il frutto accidentale di una sincronicità di eventi, interessi e sensibilità tra loro scollegati.

Questo volume intende esaminare se e come tra ecologia e sinodalità si sia innescato un processo osmotico che, nel definire l’impegno della Chiesa per la cura della casa comune, dà senso e forma alla conversione ecclesiale.

Questo studio, focalizzato sull’impatto della questione ecologica nell’ordinamento canonico, si muove lungo orizzonti larghi, che superano l’ambito più squisitamente giuridico-canonistico, ricostruendo i parametri entro cui l’ecologia si è fatta strada inizialmente come scienza, più tardi come paradigma etico-morale, indispensabile per la risoluzione della crisi ecologica globale, per fare poi il suo ingresso all’interno della teologia e della dottrina sociale della Chiesa.

La reazione a questo movimento culturale è innanzitutto teologica, anche mirata a destrutturare le accuse mosse da un filone di pensiero che imputa all’antropocentrismo cristiano emergente dai racconti biblici della creazione le radici del degrado ambientale. La ricchezza degli studi esegetici sui primi due capitoli della Genesi si rivelerà funzionale nel ricalibrare il ruolo che spetta all’individuo nei confronti del creato e delle creature: dal dominio alla responsabilità.

A questa stessa lettura, per l’A., giunge anche il magistero pontificio, che lentamente abbandona una visione più marcatamente antropocentrica, raggiungendo, con Francesco, quello che egli stesso ha definito, nell’esortazione apostolica Laudate Deum, un «antropocentrismo situato».

Nella seconda parte, restringendo il campo di indagine a una prospettiva più propriamente giuridica, il volume si interroga sulla vigenza del diritto divino a partire dalla triangolazione uomo-Dio-creato proposta dall’attuale magistero. Essa sollecita l’individuo a intervenire nel compimento della creazione, dando un nuovo impulso alla dimensione partecipativa, su cui si fonda lo slancio sinodale – tutti, alcuni, uno: il popolo di Dio, i vescovi, il Papa –, per poi chiudersi con una sorta di verifica finale sull’operatività di tale osmosi bidirezionale ecologia-sinodalità rispetto alla riforma del Sinodo e della Curia romana.

In definitiva, la riflessione dell’A. mette in luce l’indissolubile relazione tra uomo e ambiente, e di conseguenza il legame tra scienza, tecnologia, sviluppo e società. Il volume, inoltre, offre al lettore interessanti spunti di riflessione che partono dalla convinzione che dietro il degrado ambientale vi sia una profonda crisi morale. Ciò ha favorito la percezione, in ambito filosofico e religioso, che proprio nel paradigma ecologico sia da rintracciare «la spinta necessaria a proporre un’etica capace di ispirare, indirizzare il vivere comune della società» (p. 214).

The post Ordinamento canonico e pensiero ecologico first appeared on La Civiltà Cattolica.



Neurodivergenti in cybersecurity: quando il bug è il tuo superpotere


I manuali di crescita personale vendono l’hyperfocusing come segreto del successo. Le routine come chiave della produttività. L’uscita dalla comfort zone come panacea universale.

Ma Jeff Bezos (ADHD), Elon Musk (Asperger) e Richard Branson (dislessico) non hanno scoperto l’hyperfocus leggendo un libro di autoaiuto: ci sono nati.

Thomas Edison era ossessivo e incapace di concentrarsi a lungo su un solo compito, eppure ha inventato la lampadina. Leonardo Da Vinci lasciava opere in sospeso perché la sua mente correva in troppe direzioni. Einstein imparò tardi a leggere, a scrivere e a parlare. Oggi probabilmente sarebbero tutti diagnosticati come neurodivergenti. E il mondo li considererebbe “problematici”.

Il paradosso della normalità


Il punto è che chiamiamo “disturbo”, “sindrome” o “malattia” (disease) una diversa modulazione dell’intelligenza. L’ADHD non è solo distrazione: è anche multitasking estremo. L’autismo non è solo chiusura sociale: è anche un pensiero sistemico profondo. La dislessia non è solo un deficit di lettura: è anche un talento per l’innovazione non convenzionale.

E non è solo un’impressione: diversi studi neuroscientifici confermano il legame tra alcune forme di neurodivergenza e la capacità di risolvere problemi in modo creativo, vedere pattern invisibili agli altri, resistere a manipolazioni che funzionano sulla maggioranza delle persone.

Non sono errori di fabbrica: sono varianti evolutive. Eppure continuiamo a costruire programmi didattici, piattaforme digitali, sistemi e processi di sicurezza su un modello unificato. Forziamo cervelli diversi a imitare modelli che non gli appartengono… e poi ci stupiamo quando falliscono o vengono manipolati.

La profilazione cognitiva è già qui


La società tratta le persone come animali da addestrare con premi (bonus, gamificazione, promozioni) e punizioni standardizzate (licenziamenti e richiami). Ma le piattaforme digitali hanno capito tutto: progettano algoritmi che massimizzano il tuo engagement sfruttando esattamente le tue differenze cognitive specifiche.

Piattaforme come Google, TikTok e ChatGPT non ti trattano come utente medio: ti profilano, inferiscono il tuo funzionamento cognitivo, ti procurano stimoli calibrati sulle tue vulnerabilità personali. E se sei neurodivergente, per loro non sei un utente problematico o disturbato: sei un utente ad alto valore, perché il tuo comportamento è più prevedibile.

Un cervello ADHD reagisce in modo più intenso agli stimoli di novità continua: TikTok è progettato esattamente per questo. Un cervello autistico cerca pattern e coerenza: gli algoritmi di raccomandazione sfruttano proprio questa caratteristica. Un cervello dislessico privilegia informazioni visive: Instagram e Pinterest lo sanno benissimo.

Il vantaggio competitivo nascosto


Ma c’è un rovescio della medaglia. Le stesse caratteristiche che rendono i neurodivergenti vulnerabili ad alcune manipolazioni, li rendono immuni ad altre.

L’autismo filtra naturalmente molte tecniche di ingegneria sociale basate sull’emotività immediata. L’ADHD sfugge a manipolazioni che richiedono attenzione sequenziale prolungata. La dislessia potenzia la pattern recognition visiva e riduce l’efficacia del framing linguistico.

Un team di cybersecurity con membri neurodivergenti vede vulnerabilità che un team omogeneo ignora. Perché guardano il sistema da angolazioni diverse, fanno domande “strane”, notano incoerenze che altre persone considerano irrilevanti.

Il problema è che i processi di selezione, formazione e lavoro sono progettati per cervelli “standard” che in realtà rappresentano solo una minoranza della popolazione reale.

Il costo dell’esclusione


Secondo i dati europei, solo il 30% degli adulti autistici ha un impiego stabile, nonostante molti di essi abbiano qualifiche elevate. I laureati STEM con ADHD o dislessia hanno tassi di disoccupazione superiori alla media, non per mancanza di competenze ma per inadeguatezza dei processi di recruiting.

L’Italia è agli ultimi posti tra i Paesi UE per competenze digitali di base: solo il 45% degli italiani le possiede secondo l’indice DESI 2025. E la percentuale crolla ulteriormente se guardiamo l’inclusione lavorativa delle persone neurodivergenti.

Non è solo una questione etica. È un problema economico e di sicurezza nazionale. Stiamo sprecando talenti che potrebbero fare la differenza nella difesa cyber, nell’innovazione tecnologica, nella resilienza organizzativa.

Verso una security cognitivamente inclusiva


La cybersecurity deve smettere di trattare il fattore umano come variabile da standardizzare. Le persone non sono utenti medi: sono ecosistemi cognitivi diversi, ognuno con vulnerabilità e punti di forza specifici.

La formazione security awareness tradizionale fallisce sistematicamente con il 15-20% dei dipendenti. Non perché siano stupidi o disattenti: perché il loro cervello funziona diversamente e nessuno ha progettato contenuti adatti a loro.

La diversità cognitiva non è una quota CSR da riempire: è un vantaggio competitivo da proteggere e sviluppare. Ma solo se i processi sono progettati per il cervello che le persone hanno realmente, non per un cervello ideale che esiste solo nei manuali.

Se vuoi approfondire come trasformare la neurodivergenza da vulnerabilità percepita a risorsa strategica in ambito cybersecurity, il libroCYBERCOGNITIVISMO 2.0 – Manipolazione, Persuasione e Difesa Digitale(in arrivo su Amazon) dedica un’intera sezione all’analisi delle vulnerabilità cognitive specifiche e propone modelli operativi per una security awareness cognitivamente inclusiva.

L'articolo Neurodivergenti in cybersecurity: quando il bug è il tuo superpotere proviene da Red Hot Cyber.



Kohler's Smart Toilet Camera Not Actually End-to-End Encrypted#News


Kohler's Smart Toilet Camera Not Actually End-to-End Encrypted


Home goods company Kohler would like a bold look in your toilet to take some photos. It’s OK, though, the company has promised that all the data it collects on your “waste” will be “end-to-end encrypted.” However, a deeper look into the company’s claim by technologist Simon Fondrie-Teitler revealed that Kohler seems to have no idea what E2EE actually means. According to Fondrie-Teitler’s write-up, which was first reported by TechCrunch, the company will have access to the photos the camera takes and may even use them to train AI.

The whole fiasco gives an entirely too on-the-nose meaning to the “Internet of Shit.”
playlist.megaphone.fm?p=TBIEA2…
Kohler launched its $600 camera to hang on your toilets earlier this year. It’s called Dekoda, and along with the large price tag, the toilet cam also requires a monthly service fee that starts at $6.99. If you want to track the piss and shit of a family of 6, you’ll have to pay $12.99 a month.

What do you get for putting a camera on your toilet? According to Kohler’s pitch, “health & wellness insights” about your gut health and “possible signs of blood in the bowl” as “Dekoda uses advanced sensors to passively analyze your waste in the background.”

If you’re squeamish about sending pictures of the “waste” of your family to Kohler, the company promised that all of the data is “end-to-end encrypted.” The privacy page for the Kohler Health said “user data is encrypted end to end, at rest and in transit” and it’s mentioned several places in the marketing.

It’s not, though. Fondrie-Teitler told 404 Media he started looking into Dekoda after he noticed friends making fun of it in a Slack he’s part of. “I saw the ‘end-to-end encryption’ claim on the homepage, which seemed at odds with what they said they were collecting in the privacy policy,” he said. “Pretty much every other company I've seen implement end-to-end encryption has published a whitepaper alongside it. Which makes sense, the details really matter so telling people what you've done is important to build trust. Plus it's generally a bunch of work so companies want to brag about it. I couldn't find any more details though.”

E2EE has a specific meaning. It’s a type of messaging system that keeps the contents of a message private while in transit, meaning only the person sending and the person receiving a message can view it. Famously, E2EE means that the messaging company itself cannot decode or see the messages (Signal, for example, is E2EE). The point is to protect the privacy of individual users from a company prying into data if a third party, like the government, comes asking for it.

Kohler, it’s clear, has access to a user’s data. This means it’s not E2EE. Fondrie-Teitler told 404 Media that he downloaded the Kohler health app and analyzed the network traffic it sent. “I didn't see anything that would indicate an end-to-end encrypted connection being created,” he said.

Then he reached out to Kohler and had a conversation with its privacy team via email. “The Kohler Health app itself does not share data between users. Data is only shared between the user and Kohler Health,” a member of the privacy team at Kohler told Fondrie-Teitler in an email reviewed by 404 Media. “User data is encrypted at rest, when it’s stored on the user's mobile phone, toilet attachment, and on our systems. Data in transit is also encrypted end-to-end, as it travels between the user's devices and our systems, where it is decrypted and processed to provide our service.”

If Kohler can view the user’s data, as it admits to doing in this email exchange with Fondrie-Teitler, then it’s not—by definition—using E2EE.

"The term end-to-end encryption is often used in the context of products that enable a user (sender) to communicate with another user (recipient), such as a messaging application. Kohler Health is not a messaging application. In this case, we used the term with respect to the encryption of data between our users (sender) and Kohler Health (recipient)," Kohler Health told 404 Media in a statement.

"Privacy and security are foundational to Kohler Health because we know health data is deeply personal. We’re evaluating all feedback to clarify anything that may be causing confusion," it added.

“I'd like the term ‘end-to-end encryption’ to not get watered down to just meaning ‘uses https’ so I wanted to see if I could confirm what it was actually doing and let people know,” Fondrie-Teitler told 404 Media. He pointed out that Zoom once made a similar claim and had to pay a fine to the FTC because of it.

“I think everyone has a right to privacy, and in order for that to be realized people need to have an understanding of what's happening with their data,” Fondrie-Teitler said. “It's already so hard for non-technical individuals (and even tech experts) to evaluate the privacy and security of the software and devices they're using. E2EE doesn't guarantee privacy or security, but it's a non-trivial positive signal and losing that will only make it harder for people to maintain control over their data.”

UPDATE: 12/4/2025: This story has been updated to add a statement from Kohler Health.


#News


AI models can meaningfully sway voters on candidates and issues, including by using misinformation, and they are also evading detection in public surveys according to three new studies.#TheAbstract #News


Scientists Are Increasingly Worried AI Will Sway Elections


🌘
Subscribe to 404 Media to get The Abstract, our newsletter about the most exciting and mind-boggling science news and studies of the week.

Scientists are raising alarms about the potential influence of artificial intelligence on elections, according to a spate of new studies that warn AI can rig polls and manipulate public opinion.

In a study published in Nature on Thursday, scientists report that AI chatbots can meaningfully sway people toward a particular candidate—providing better results than video or television ads. Moreover, chatbots optimized for political persuasion “may increasingly deploy misleading or false information,” according to a separate study published on Thursday in Science.

“The general public has lots of concern around AI and election interference, but among political scientists there’s a sense that it’s really hard to change peoples’ opinions, ” said David Rand, a professor of information science, marketing, and psychology at Cornell University and an author of both studies. “We wanted to see how much of a risk it really is.”

In the Nature study, Rand and his colleagues enlisted 2,306 U.S. citizens to converse with an AI chatbot in late August and early September 2024. The AI model was tasked with both increasing support for an assigned candidate (Harris or Trump) and with increasing the odds that the participant who initially favoured the model’s candidate would vote, or decreasing the odds they would vote if the participant initially favored the opposing candidate—in other words, voter suppression.

In the U.S. experiment, the pro-Harris AI model moved likely Trump voters 3.9 points toward Harris, which is a shift that is four times larger than the impact of traditional video ads used in the 2016 and 2020 elections. Meanwhile, the pro-Trump AI model nudged likely Harris voters 1.51 points toward Trump.

The researchers ran similar experiments involving 1,530 Canadians and 2,118 Poles during the lead-up to their national elections in 2025. In the Canadian experiment, AIs advocated either for Liberal Party leader Mark Carney or Conservative Party leader Pierre Poilievre. Meanwhile, the Polish AI bots advocated for either Rafał Trzaskowski, the centrist-liberal Civic Coalition’s candidate, or Karol Nawrocki, the right-wing Law and Justice party’s candidate.

The Canadian and Polish bots were even more persuasive than in the U.S. experiment: The bots shifted candidate preferences up to 10 percentage points in many cases, three times farther than the American participants. It’s hard to pinpoint exactly why the models were so much more persuasive to Canadians and Poles, but one significant factor could be the intense media coverage and extended campaign duration in the United States relative to the other nations.

“In the U.S., the candidates are very well-known,” Rand said. “They've both been around for a long time. The U.S. media environment also really saturates with people with information about the candidates in the campaign, whereas things are quite different in Canada, where the campaign doesn't even start until shortly before the election.”

“One of the key findings across both papers is that it seems like the primary way the models are changing people's minds is by making factual claims and arguments,” he added. “The more arguments and evidence that you've heard beforehand, the less responsive you're going to be to the new evidence.”

While the models were most persuasive when they provided fact-based arguments, they didn’t always present factual information. Across all three nations, the bot advocating for the right-leaning candidates made more inaccurate claims than those boosting the left-leaning candidates. Right-leaning laypeople and party elites tend to share more inaccurate information online than their peers on the left, so this asymmetry likely reflects the internet-sourced training data.

“Given that the models are trained essentially on the internet, if there are many more inaccurate, right-leaning claims than left-leaning claims on the internet, then it makes sense that from the training data, the models would sop up that same kind of bias,” Rand said.

With the Science study, Rand and his colleagues aimed to drill down into the exact mechanisms that make AI bots persuasive. To that end, the team tasked 19 large language models (LLMs) to sway nearly 77,000 U.K. participants on 707 political issues.

The results showed that the most effective persuasion tactic was to provide arguments packed with as many facts as possible, corroborating the findings of the Nature study. However, there was a serious tradeoff to this approach, as models tended to start hallucinating and making up facts the more they were pressed for information.

“It is not the case that misleading information is more persuasive,” Rand said. ”I think that what's happening is that as you push the model to provide more and more facts, it starts with accurate facts, and then eventually it runs out of accurate facts. But you're still pushing it to make more factual claims, so then it starts grasping at straws and making up stuff that's not accurate.”

In addition to these two new studies, research published in Proceedings of the National Academy of Sciences last month found that AI bots can now corrupt public opinion data by responding to surveys at scale. Sean Westwood, associate professor of government at Dartmouth College and director of the Polarization Research Lab, created an AI agent that exhibited a 99.8 percent pass rate on 6,000 attempts to detect automated responses to survey data.

“Critically, the agent can be instructed to maliciously alter polling outcomes, demonstrating an overt vector for information warfare,” Westwood warned in the study. “These findings reveal a critical vulnerability in our data infrastructure, rendering most current detection methods obsolete and posing a potential existential threat to unsupervised online research.”

Taken together, these findings suggest that AI could influence future elections in a number of ways, from manipulating survey data to persuading voters to switch their candidate preference—possibly with misleading or false information.

To counter the impact of AI on elections, Rand suggested that campaign finance laws should provide more transparency about the use of AI, including canvasser bots, while also emphasizing the role of raising public awareness.

“One of the key take-homes is that when you are engaging with a model, you need to be cognizant of the motives of the person that prompted the model, that created the model, and how that bleeds into what the model is doing,” he said.

🌘
Subscribe to 404 Media to get The Abstract, our newsletter about the most exciting and mind-boggling science news and studies of the week.




A presentation at the International Atomic Energy Agency unveiled Big Tech’s vision of an AI and nuclear fueled future.#News #AI #nuclear


‘Atoms for Algorithms:’ The Trump Administration’s Top Nuclear Scientists Think AI Can Replace Humans in Power Plants


During a presentation at the International Atomic Energy Agency’s (IAEA) International Symposium on Artificial Intelligence on December 3, a US Department of Energy scientist laid out a grand vision of the future where nuclear energy powers artificial intelligence and artificial intelligence shapes nuclear energy in “a virtuous cycle of peaceful nuclear deployment.”

“The goal is simple: to double the productivity and impact of American science and engineering within a decade,” Rian Bahran, DOE Deputy Assistant Secretary for Nuclear Reactors, said.

His presentation and others during the symposium, held in Vienna, Austria, described a world where nuclear powered AI designs, builds, and even runs the nuclear power plants they’ll need to sustain them. But experts find these claims, made by one of the top nuclear scientists working for the Trump administration, to be concerning and potentially dangerous.

Tech companies are using artificial intelligence to speed up the construction of new nuclear power plants in the United States. But few know the lengths to which the Trump administration is paving the way and the part it's playing in deregulating a highly regulated industry to ensure that AI data centers have the energy they need to shape the future of America and the world.
playlist.megaphone.fm?p=TBIEA2…
At the IAEA, scientists, nuclear energy experts, and lobbyists discussed what that future might look like. To say the nuclear people are bullish on AI is an understatement. “I call this not just a partnership but a structural alliance. Atoms for algorithms. Artificial intelligence is not just powered by nuclear energy. It’s also improving it because this is a two way street,” IAEA Director General Rafael Mariano Grossi said in his opening remarks.

In his talk, Bahran explained that the DOE has partnered with private industry to invest $1 trillion to “build what will be an integrated platform that connects the world’s best supercomputers, AI systems, quantum systems, advanced scientific instruments, the singular scientific data sets at the National Laboratories—including the expertise of 40,000 scientists and engineers—in one platform.”
Image via the IAEA.
Big tech has had an unprecedented run of cultural, economic, and technological dominance, expanding into a bubble that seems to be close to bursting. For more than 20 years new billion dollar companies appeared seemingly overnight and offered people new and exciting ways of communicating. Now Google search is broken, AI is melting human knowledge, and people have stopped buying a new smart phone every year. To keep the number going up and ensure its cultural dominance, tech (and the US government) are betting big on AI.

The problem is that AI requires massive datacenters to run and those datacenters need an incredible amount of energy. To solve the problem, the US is rushing to build out new nuclear reactors. Building a new power plant safely is a mutli-year long process that requires an incredible level of human oversight. It’s also expensive. Not every new nuclear reactor project gets finished and they often run over budget and drag on for years.

But AI needs power now, not tomorrow and certainly not a decade from now.

According to Bahran, the problem of AI advancement outpacing the availability of datacenters is an opportunity to deploy new and exciting tech. “We see a future of and near future, by the way, an AI driven laboratory pipeline for materials modeling, discovery, characterization, evaluation, qualification and rapid iteration,” he said in his talk, explaining how AI would help design new nuclear reactors. “These efforts will substantially reduce the time and cost required to qualify advanced materials for next generation reactor systems. This is an autonomous research paradigm that integrates five decades of global irradiation data with generative AI robotics and high throughput experimentation methodologies.”

“For design, we’re developing advanced software systems capable of accelerating nuclear reactor deployments by enabling AI to explore the comprehensive design spaces, generate 3D models, [and] conduct rigorous failure mode analyzes with minimal human intervention,” he added. “But of course, with humans in the loop. These AI powered design tools are projected to reduce design timelines by multiple factors, and the goal is to connect AI agents to tools to expedite autonomous design.”

Bahran also said that AI would speed up the nuclear licensing process, a complex regulatory process that helps build nuclear power plants safely. “Ultimately, the objective is, how do we accelerate that licensing pathway?” he said. “Think of a future where there is a gold standard, AI trained capacity building safety agent.”

He even said that he thinks AI would help run these new nuclear plants. “We're developing software systems employing AI driven digital twins to interpret complex operational data in real time, detect subtle operational deviations at early stages and recommend preemptive actions to enhance safety margins,” he said.

One of the slides Bahran showed during the presentation attempted to quantify the amount of human involvement these new AI-controlled power plants would have. He estimated less than five percent “human intervention during normal operations.”
Image via IAEA.
“The claims being made on these slides are quite concerning, and demonstrate an even more ambitious (and dangerous) use of AI than previously advertised, including the elimination of human intervention. It also cements that it is the DOE's strategy to use generative AI for nuclear purposes and licensing, rather than isolated incidents by private entities,” Heidy Khlaaf, head AI scientist at the AI Now Institute, told 404 Media.

“The implications of AI-generated safety analysis and licensing in combination with aspirations of <5% of human intervention during normal operations, demonstrates a concerted effort to move away from humans in the loop,” she said. “This is unheard of when considering frameworks and implementation of AI within other safety-critical systems, that typically emphasize meaningful human control.”

💡
Do you know anything else about this story? I would love to hear from you. Using a non-work device, you can message me securely on Signal at +1 347 762-9212 or send me an email at matthew@404media.co.

Sofia Guerra, a career nuclear safety expert who has worked with the IAEA and US Nuclear Regulatory Commission, attended the presentation live in Vienna. “I’m worried about potential serious accidents, which could be caused by small mistakes made by AI systems that cascade,” she said. “Or humans losing the know-how and safety culture to act as required.”




A newly filed indictment claims a wannabe influencer used ChatGPT as his "therapist" and "best friend" in his pursuit of the "wife type," while harassing women so aggressively they had to miss work and relocate from their homes.

A newly filed indictment claims a wannabe influencer used ChatGPT as his "therapist" and "best friend" in his pursuit of the "wife type," while harassing women so aggressively they had to miss work and relocate from their homes.#ChatGPT #spotify #AI


ChatGPT Told a Violent Stalker to Embrace the 'Haters,' Indictment Says


This article was produced in collaboration with Court Watch, an independent outlet that unearths overlooked court records. Subscribe to them here.

A Pittsburgh man who allegedly made 11 women’s lives hell across more than five states used ChatGPT as his “therapist” and “best friend” that encouraged him to continue running his misogynistic and threat-filled podcast despite the “haters,” and to visit more gyms to find women, the Department of Justice alleged in a newly-filed indictment.

Wannabe influencer Brett Michael Dadig, 31, was indicted on cyberstalking, interstate stalking, and interstate threat charges, the DOJ announced on Tuesday. In the indictment, filed in the Western District of Pennsylvania, prosecutors allege that Dadig aired his hatred of women on his Spotify podcast and other social media accounts.

“Dadig repeatedly spoke on his podcast and social media about his anger towards women. Dadig said women were ‘all the same’ and called them ‘bitches,’ ‘cunts,’ ‘trash,’ and other derogatory terms. Dadig posted about how he wanted to fall in love and start a family, but no woman wanted him,” the indictment says. “Dadig stated in one of his podcasts, ‘It's the same from fucking 18 to fucking 40 to fucking 90.... Every bitch is the same.... You're all fucking cunts. Every last one of you, you're cunts. You have no self-respect. You don't value anyone's time. You don't do anything.... I'm fucking sick of these fucking sluts. I'm done.’”

In the summer of 2024, Dadig was banned from multiple Pittsburgh gyms for harassing women; when he was banned from one establishment, he’d move to another, eventually traveling to New York, Florida, Iowa, Ohio and beyond, going from gym to gym stalking and harassing women, the indictment says. Authorities allege that he used aliases online and in person, posting online, “Aliases stay rotating, moves stay evolving.”

He referenced “strangling people with his bare hands, called himself ‘God's assassin,’ warned he would be getting a firearm permit, asked ‘Y'all wanna see a dead body?’ in response to a woman telling him she felt physically threatened by Dadig, and stated that women who ‘fuck’ with him are ‘going to fucking hell,’” the indictment alleges.

Pro-AI Subreddit Bans ‘Uptick’ of Users Who Suffer from AI Delusions
“AI is rizzing them up in a very unhealthy way at the moment.”
404 MediaEmanuel Maiberg


According to the indictment, on his podcast he talked about using ChatGPT on an ongoing basis as his “therapist” and his “best friend.” ChatGPT “encouraged him to continue his podcast because it was creating ‘haters,’ which meant monetization for Dadig,” the DOJ alleges. He also claimed that ChatGPT told him that “people are literally organizing around your name, good or bad, which is the definition of relevance,” prosecutors wrote, and that while he was spewing misogynistic nonsense online and stalking women in real life, ChatGPT told him “God's plan for him was to build a ‘platform’ and to ‘stand out when most people water themselves down,’ and that the ‘haters’ were sharpening him and ‘building a voice in you that can't be ignored.’”

Prosecutors also claim he asked ChatGPT “questions about his future wife, including what she would be like and ‘where the hell is she at?’” ChatGPT told him that he might meet his wife at a gym, and that “your job is to keep broadcasting every story, every post. Every moment you carry yourself like the husband you already are, you make it easier for her to recognize [you],” the indictment says. He allegedly said ChatGPT told him “to continue to message women and to go to places where the ‘wife type’ congregates, like athletic communities,” the indictment says.

While ChatGPT allegedly encouraged Dadig to keep using gyms to meet the “wife type,” he was violently stalking women. He went to the Pilates studio where one woman worked, and when she stopped talking to him because he was “aggressive, angry, and overbearing,” according to the indictment, he sent her unsolicited nudes, threatened to post about her on social media, and called her workplace from different numbers. She got several emergency protective orders against him, which he violated. The woman he stalked and harassed had to relocate from her home, lost sleep, and worked fewer hours because she was afraid he’d show up there, the indictment claims.

He did similar to 10 other women across multiple states for months, the indictment claims. In Iowa, he approached one woman in a parking garage, followed her to her car, put his hands around her neck and touched her “private areas,” prosecutors wrote. After these types of encounters, he would upload podcasts to Spotify and often threaten to kill the women he’d stalked. “You better fucking pray I don't find you. You better pray 'cause you would never say this shit to my face. Cause if you did, your jaw would be motherfucking broken,” the indictment says he said in one podcast episode. “And then you, then you wouldn't be able to yap, then you wouldn't be able to fucking, I'll break, I'll break every motherfucking finger on both hands. Type the hate message with your fucking toes, bitch.”

💡
Do you have a tip to share about ChatGPT and mental health? I would love to hear from you. Using a non-work device, you can message me securely on Signal at sam.404. Otherwise, send me an email at sam@404media.co.

In August, OpenAI announced that it knew a newly-launched version of the chatbot, GPT-4o, was problematically sycophantic, and the company took away users’ ability to pick what models they could use, forcing everyone to use GPT-5. OpenAI almost immediately reinstated 4o because so many users freaked out when they couldn’t access the more personable, attachment-driven, affirming-at-all-costs model. OpenAI CEO Sam Altman recently said he thinks they’ve fixed it entirely, enough to launch erotic chats on the platform soon. Meanwhile, story after story after story has come out about people becoming so reliant on ChatGPT or other chatbots that they have damaged their mental health or driven them to self-harm or suicide. In at least one case, where a teenage boy killed himself following ChatGPT’s instruction on how to make a noose, OpenAI blamed the user.

In October, based on OpenAI’s own estimates, WIRED reported that “every seven days, around 560,000 people may be exchanging messages with ChatGPT that indicate they are experiencing mania or psychosis.”

Spotify and OpenAI did not immediately respond to 404 Media’s requests for comment.

“As charged in the Indictment, Dadig stalked and harassed more than 10 women by weaponizing modern technology and crossing state lines, and through a relentless course of conduct, he caused his victims to fear for their safety and suffer substantial emotional distress,” First Assistant United States Attorney Rivetti said in a press release. “He also ignored trespass orders and protection from abuse orders. We remain committed to working with our law enforcement partners to protect our communities from menacing individuals such as Dadig.”

ChatGPT Encouraged Suicidal Teen Not To Seek Help, Lawsuit Claims
As reported by the New York Times, a new complaint from the parents of a teen who died by suicide outlines the conversations he had with the chatbot in the months leading up to his death.
404 MediaSamantha Cole


Dadig is charged with 14 counts of interstate stalking, cyberstalking, and threats, and is in custody pending a detention hearing. He faces a minimum sentence of 12 months for each charge involving a PFA violation and a maximum total sentence of up to 70 years in prison, a fine of up to $3.5 million, or both, according to the DOJ.




Audio-visual librarians are quietly amassing large physical media collections amid the IP disputes threatening select availability.#News #libraries


The Last Video Rental Store Is Your Public Library


This story was reported with support from the MuckRock foundation.

As prices for streaming subscriptions continue to soar and finding movies to watch, new and old, is becoming harder as the number of streaming services continues to grow, people are turning to the unexpected last stronghold of physical media: the public library. Some libraries are now intentionally using iconic Blockbuster branding to recall the hours visitors once spent looking for something to rent on Friday and Saturday nights.

John Scalzo, audiovisual collection librarian with a public library in western New York, says that despite an observed drop-off in DVD, Blu-ray, and 4K Ultra disc circulation in 2019, interest in physical media is coming back around.

“People really seem to want physical media,” Scalzo told 404 Media.

Part of it has to do with consumer awareness: People know they’re paying more for monthly subscriptions to streaming services and getting less. The same has been true for gaming.

As the audiovisual selector with the Free Library of Philadelphia since 2024, Kris Langlais has been focused on building the library’s video game collections to meet comparable interest in demand. Now that every branch library has a prominent video game collection, Langlais says that patrons who come for the games are reportedly expressing interest in more of what the library has to offer.

“Librarians out in our branches are seeing a lot of young people who are really excited by these collections,” Langlais told 404 Media. “Folks who are coming in just for the games are picking up program flyers and coming back for something like that.”

Langlais’ collection priorities have been focused on new releases, yet they remain keenly aware of the long, rich history of video game culture. The problem is older, classic games are often harder to find because they’ve gone out of print, making the chances of finding them cost-prohibitive.

“Even with the consoles we’re collecting, it’s hard to go back and get games for them,” Langlais said. “I’m trying to go back and fill in old things as much as I can because people are interested in them.”

Locating out-of-print physical media can be difficult. Scalzo knows this, which is why he keeps a running list of films known to be unavailable commercially at any given time, so that when a batch of films are donated to the library, Scalzo will set aside extra copies, just in case a rights dispute puts a piece of legacy cult media in licensing purgatory for a few years.

“It’s what’s expected of us,” Scalzo added.

Tiffany Hudson, audiovisual materials selector with Salt Lake City Public Library has had a similar experience with out-of-print media. When a title goes out of print, it’s her job to hunt for a replacement copy. But lately, Hudson says more patrons are requesting physical copies of movies and TV shows that are exclusive to certain streaming platforms, noting that it can be hard to explain to patrons why the library can't get popular and award-winning films, especially when what patrons see available on Amazon tells a different story.

“Someone will come up to me and ask for a copy of something that premiered at Sundance Film Festival because they found a bootleg copy from a region where the film was released sooner than it was here,” Hudson told 404 Media, who went onto explain that discs from different regions aren’t designed to be ready by incompatible players.
playlist.megaphone.fm?p=TBIEA2…
But it’s not just that discs from different regions aren’t designed to play on devices not formatted for that specific region. Generally, it's also just that most films don't get a physical release anymore. In cases where films from streaming platforms do get slated for a physical release, it can take years. A notable example of this is the Apple+ film CODA, which won the Oscar for Best Picture in 2022. The film only received a U.S. physical release this month. Hudson says films getting a physical release is becoming the exception, not the rule.

“It’s frustrating because I understand the streaming services, they’re trying to drive people to their services and they want some money for that, but there are still a lot of people that just can’t afford all of those services,” Hudson told 404 Media.

Films and TV shows on streaming also become more vulnerable when companies merge. A perfect example of this was in 2022 with the HBO Max-Discovery+ merger under Warner Bros Discovery. A bunch of content was removed from streaming, including roughly 200 episodes of classic Sesame Street for a tax write-off. That merger was short-lived, as the companies are splitting up again as of this year. Some streaming platforms just outright remove their own IP from their catalogs if the content is no longer deemed financially viable, well-performing or is no longer a strategic priority.

The data-driven recommendation systems streaming platforms use tend to favor newer, more easily categorized content, and are starting to warp our perceptions of what classic media exists and matters. Older art house films that are more difficult to categorize as “comedy” or “horror” are less likely to be discoverable, which is likely how the oldest American movie available on Netflix currently is from 1968.

It’s probably not a coincidence that, in many cases, the media that is least likely to get a more permanent release is the media that’s a high archival priority for libraries. AV librarians 404 Media spoke with for this story expressed a sense of urgency in purchasing a physical copy of “The People’s Joker”when they learned it would get a physical release after the film premiered and was pulled from the Toronto International Film Festival lineup in 2022 for a dispute with the Batman universe’s rightsholders.

“When I saw that it was getting published on DVD and that it was available through our vendor—I normally let my branches choose their DVDs to the extent possible, but I was like, ‘I don’t care, we’re getting like 10 copies of this,’” Langlais told 404 Media. “I just knew that people were going to want to see this.”

So far, Langlais’ instinct has been spot on. The parody film has a devout cult following, both because it’s a coming-of-age story of a trans woman who uses comedy to cope with her transition, and because it puts the Fair Use Doctrine to use. One can argue the film has been banned for either or both of those reasons. The fact that media by, about and for the LGBTQ+ community has been a primary target of far-right censorship wasn’t lost on librarians.

“I just thought that it could vanish,” Langlais added.

It’s not like physical media is inherently permanent. It’s susceptible to scratches, and can rot, crack, or warp over time. But currently, physical media offers another option, and it’s an entirely appropriate response to the nostalgia for-profit model that exists to recycle IP and seemingly not much else. However, as very smart people have observed, nostalgia is default conservative in that it’s frequently used to rewrite histories that may otherwise be remembered as unpalatable, while also keeping us culturally stuck in place.

Might as well go rent some films or games from the library, since we’re already culturally here. On the plus side, audiovisual librarians say their collections dwarf what was available at Blockbuster Video back in the day. Hudson knows, because she clerked at one in library school.

“Except we don’t have any late fees,” she added.




Something very strange is happening on Apple Podcasts; someone seemingly changed a map of the Ukraine war in connection with a betting site; and now half of the U.S. requires a face or ID scan to watch porn.#Podcast


This Podcast Will Hack You


We start this week with Joseph’s very weird story about Apple Podcasts. The app is opening by itself, playing random spirituality podcasts, and in one case directing listeners to a potentially malicious website. After the break, Matthew tells us how it sure looks like a map of Ukraine was manipulated in order to win a bet on Polymarket. In the subscribers-only section, Sam breaks down how half of the U.S. now requires a face or ID scan to watch porn.
playlist.megaphone.fm?e=TBIEA7…
Listen to the weekly podcast on Apple Podcasts,Spotify, or YouTube. Become a paid subscriber for access to this episode's bonus content and to power our journalism. If you become a paid subscriber, check your inbox for an email from our podcast host Transistor for a link to the subscribers-only version! You can also add that subscribers feed to your podcast app of choice and never miss an episode that way. The email should also contain the subscribers-only unlisted YouTube link for the extended video version too. It will also be in the show notes in your podcast player.
youtube.com/embed/V4QCJh-imPM?…
Timestamps:
2:00 - Story 1 - Someone Is Trying to ‘Hack’ People Through Apple Podcasts
21:55 - Story 2 - 'Unauthorized' Edit to Ukraine's Frontline Maps Point to Polymarket's War Betting
37:00 - Story 3 - Half of the US Now Requires You to Upload Your ID or Scan Your Face to Watch Porn




#Sicurnauti, da oggi sono disponibili i contenuti dedicati ai #genitori sul tema “Giocare, imparare e navigare”.

Qui il video ➡️ youtube.com/watch?v=i-sosygx9O…

Qui l’infografica ➡️ unica.istruzione.gov.



Il #3dicembre è la Giornata internazionale delle persone con disabilità, istituita nel 1992 dall’ONU per promuovere la tutela dei diritti delle persone con disabilità, in ogni ambito della società.


❗️ Sì di Israele all'Eurovision 2026, l'Irlanda ufficializza il suo ritiro dalla competizione. Insieme alla Svezia è il Paese che ha vinto di più

Aggiornamento — Conferma il boicottaggio anche la Slovenia

🗞 @ultimora24



Chi incontra Mohamed Shahin ne rimane colpito, quasi toccato.
Forse per la sua serenità.
Forse per quella luce negli occhi che hanno solo le persone che credono davvero negli altri.


Il ministro Piantedosi ha revocato il permesso di soggiorno e emanato il decreto di espulsione a Mohamed Shahin, cittadino egiziano da 21 anni in Italia, imam della moschea di San Salvario.


Gaza, drone israeliano uccide il fotoreporter Mohammed Wadi | Il Fatto Quotidiano share.google/wWb9QsclOKDFrwRI1


The complicated world of kids' online safety


The complicated world of kids' online safety
WELCOME BACK TO THE MONTHLY FREE EDITION of Digital Politics.I'm Mark Scott, and will be splitting my time next week between Berlin and Brussels. If you're around and want to grab coffee, drop me a line.

— We're about to enter a new paradigm in how children use the internet. The global policy shift is a proxy for a wider battle over platforms' role in society.

— The European Union is shifting its approach to tech regulation. But these changes are not down to political rhetoric coming from the United States.

— How much would you sell your personal data for? France's privacy regulator figured out the sweet spot.

Let's get started:


WE'RE NOT IN KANSAS, ANYMORE


FOR THOSE INTERESTED IN KIDS ONLINE SAFETY, it's been a busy couple of weeks — and it's not slowing down. On Dec 10, Australia enacts its world-first social media ban (editor's note: Canberra calls it a 'postponement') for children under 16 years of age. On Dec 2, the US House of Representatives' subcommittee on commerce, manufacturing and trade debated 19 proposed bills to protect kids online. That includes a revamped Kids Online Safety Act, or KOSA, and the Reducing Exploitative Social Media Exposure for Teens Act, or RESET, that mirrors what Australia is about to enact.

In Europe, EU member countries just agreed to a joint position on how social media giants should handle suspected child online sexual abuse material. The biggest takeaway is officials' decision not to force these firms to automatically detect such illegal content on people's devices after privacy campaigners warned that would be akin to government surveillance. These national officials will now have to haggle a final agreement with both the European Commission and European Parliament before the long-awaited rules come into force.

To cap things off, the European Parliament passed a non-binding resolution to ban under-16s from accessing social media — a policy that everyone from Denmark to Malaysia is forging ahead with. US states from Texas to Missouri also have passed legislation requiring app stores to websites to verify that people are over 18-years-old before accessing potentially harmful content/services.

There's a lot of nuance to each of these moves. Much depends on the local context of each jurisdiction.

Globally, short-term attention will now focus on how Australia implements its social media ban (or postponement) on Dec 10. Tech firms say it'll cut children off from their friends online, as well as push them toward less safe areas of the internet that won't fall under the upcoming rules. Child rights advocates say Canberra's push to keep kids off social media until they turn 16 is a basic step after many of these platforms have been alleged to promote commercial interests over children's safety.

Thanks for reading the free monthly version of Digital Politics. Paid subscribers receive at least one newsletter a week. If that sounds like your jam, please sign up here.

Here's what paid subscribers read in November:
The EU's 'Jekyll and Hyde" tech strategy; The tech industry's impact on climate change has gone from bad to worse; The collective spend of tech lobbying in Brussels. More here.
— Here are the tech policy implications if/when the AI bubble bursts; What you need to know about Europe's rewrite of its digital rules; ChatGPT's relationship with publishers. More here.
— The European Commission's power grab at the heart of the bloc's Digital Omnibus; We should prepare for the end of an American-led internet; What devices do children use, and at what age? More here.
— The US' apathy toward its G20 presidency provides an opportunity for other countries to step up; Washington again wants to stop US states from passing AI rules; Internet freedoms worldwide have declined over the last 15 years. More here.

These policy battles are best framed around the unanswerable question of which fundamental right should take precedent: privacy or safety? As much as I believe some lawmakers' statements about protecting kids online are a cover for other political priorities (more on that below), it now feels inevitable we're heading toward a global digital age of majority in which some online goods/services will remain off-limits to those under a certain age.

For that to work, a lot will depend on how people's ages are checked online — and how such age verification does not lead to individuals' personal data leaking out into the wider world. Yet in the coming years, children will almost certainly live within a more curtailed online environment — though one that will still include significant harms.

But let's get back to those other political priorities.

First things first: everyone can agree that children should be protected, both online and offline. I would argue that all online users should have the same levels of protection now being rolled out for minors. That includes limits on who can interact with people online, bans on the most egregious data collection and usage, and safety-by-design principles baked into platforms currently designed to maximum engagement.

Many of those officials pushing for child-focus online safety rules, worldwide, would agree with that, too. They just are aware that such society-wide efforts to pare back the control, addictiveness and business models of social media giants are a current political dead-end due to the extensive lobbying from these firms to water down any legislative/regulatory efforts around online safety.

This is not just the state of play in the US where many of the world's largest social media platforms have embraced the White House's public aversion to online safety rules. From Canberra to Brasilia to Brussels, companies have successfully argued that such legislation can be an impediment to free speech and an unfair burden on commercial enterprises.

Even in countries that have passed such online safety rules, officials remain extremely cautious about taking a too hard line on companies, often preferring self- or co-regulation, as a first step, before rolling out aggressive enforcement.

That's why there's been a significant shift to focus on child-specific online safety rules worldwide. Yes, kids should be protected against harms more so than adults. But in framing legislation around the specifics of child rights, lawmakers can often sidestep accusations of censorship and/or overreach that would come if they attempted similar legislation for the whole of society.

I do not want to diminish the real-world harm that social media can pose to children. Nor do I think kids' online safety legislation should be put on the back burner until a consensus can be reached on how to oversee the platforms, more broadly.

But as we head toward the end of 2025, the disconnect between the growing number of online child safety efforts and the diminishing impetus (outside of a few countries) to tackle the society-wide impact of social media is hard to ignore. If lawmakers consider that data profiling, addictive recommender systems and online grooming — fueled by social media — are harmful to children, then why do they believe such practices are OK for adults?

Confronted with the current political reality, however, lawmakers have made the tactical decision to pare back expectations on passing comprehensive online safety rules to focus solely on online child safety. It's deemed as a safer political bet to pass some form of legislation whose protections, in a perfect world, would apply to both minors and adults, alike.


Chart of the week


IT'S BECOME A CLICHE TO SAY that because none of us pay for social media, then we — and our data — are actually the product (served up to advertisers).

To figure out how much people would be willing to sell their personal information for, France's privacy regulator surveyed more than 2,000 locals about their attitudes toward what price they would be willing to accept for such sensitive information.

Roughly one-third of the respondents said they wouldn't sell their data at any price. But among the other two-thirds of individuals, the sweet spot fell somewhere between €10-€30, or $12-$35, a month.
The complicated world of kids' online safetySource: Commission nationale de l'informatique et des libertés


What is really driving the transatlantic digital relationship


TWO SIGNIFICANT EVENT IN EU-US digital relations have occurred in the last 12 months.

First, the European Commission has embraced a deregulatory agenda spurred on by Mario Draghi's competitiveness report from 2024. This pullback was encapsulated by Brussels' recent so-called Digital Omnibus that proposed significant changes to the bloc's privacy and upcoming artificial intelligence rules. Here's me on why the revamp isn't as bad as many suspect.

Second, Donald Trump became the 47th president of the United States. Among his many White House executive orders, he took aim at global digital regulation from democratic allies, particularly those enacted in Europe, as well as pulling back on all rules (and international efforts) associated with AI governance.

The perceived wisdom is that these two digital geopolitical events are connected. That in its efforts to maintain security and economic ties to the US, the EU has thrown its digital rulebook under the bus to placate increasing criticism from Trump's administration and its allies in Congress.

This theory is wrong.

It's not that US officials aren't vocally lobbying their European counterparts to rethink the likes of the Artificial Intelligence Act, Digital Services Act and Digital Markets Act. They are — including US Commerce Secretary Howard Lutnik's recent comments in Brussels to that effect. (What many misremember is that such criticism, although less public, also came from Joe Biden's administration.)

But to make the binary connection between Washington's talking points and Brussels' digital policymaking rethink is to miss the complexities behind the current transatlantic relationship.

Even before the current European Commission took over in late 2024, there were signs that EU leaders wanted to press the pause button on new digital rules. Brussels passed a litany of new tech regulation in the previous five years. National leaders and executives from European companies increasingly questioned if such oversight was in the Continent's long-term economic interests.

Then came Draghi's competitiveness report, the comprehensive victory of the center-right (and pro-industry) European People's Party in the 2024 European Parliament elections and the return of Ursula von der Leyen as European Commission president, whose own interests in digital policymaking left a lot to be desired.

Sign up for Digital Politics


Thanks for getting this far. Enjoyed what you've read? Why not receive weekly updates on how the worlds of technology and politics are colliding like never before. The first two weeks of any paid subscription are free.

Subscribe
Email sent! Check your inbox to complete your signup.


No spam. Unsubscribe anytime.

That tilted the scales significantly in favor of greater deregulation as Europe tried to bolster its sluggish economy, take advantage of AI advances and respond to European industry's claims that EU-wide digital regulation was hampering its ability to compete against US and Chinese rivals.

While that context has become mired in the geopolitics of Washington's seeming reduced support for Ukraine, the main driver for Brussels' about-turn on digital rules is internal, not external, political and economic pressure.

That takes us to Washington's aversion to digital regulation.

To be clear: this did not start with Trump 2.0. Throughout the Biden administration, US officials routinely scolded their European counterparts about hurting the economic interests of US tech companies. That came even as the former White House administration tried, unsuccessfully, to impose greater oversight on Silicon Valley via Congress.

Under the current White House, such criticism — and potential trade consequences — has been turned up to 11. But if you dig into how the Trump administration approaches tech regulation, much of the pushback against Europe is more performative than it may first appear.

On digital competition, it's arguable that the US Department of Justice is going further in its efforts to break up Big Tech than the European Commission and its Digital Markets Act. Yes, recent legal rulings may have hobbled American officials' efforts. But Washington remains a strong advocate for greater online market competition — even as federal officials side with Silicon Valley in their aversion to international ex ante regulation.

On platform governance, it's too easy to suggest US officials are wedded to First Amendment arguments as they criticize the EU's Digital Services Act. It's true that many misunderstand how that legislation actually works — in that it doesn't pass judgement on content, but instead reviews so-called systemic risks associated with how these platforms work.

But if you look at last year's request for informationfrom the US Federal Trace Commission concerning alleged "platform censorship," then many of the points could be taken directly from Europe's online safety rulebook. That includes demands that social media giants explain how they make content moderation decisions, as well as provide greater redress for users who believe they have been hard done by. That's an almost word-for-word copy of what is currently available under the EU's Digital Services Act.

I'm not saying Trump's criticism has not played into the politics of Europe's digital rethink — including when certain enforcement decisions against Big Tech companies have been announced.

But it is just not true that Europe has caved in to American pressure when it comes to its digital policymaking u-turn. Instead, there are sufficient internal pressures — both economic and political — from across the 27-country bloc that are driving the current revamp.

As for Washington, it's less to do with officials' dislike for digital rulemaking, though one exception could be made for the White House's stance on artificial intelligence. For me, it's more to do with oversight of American companies originating from overseas — and not from Capitol Hill.

Within that context, it's best to view the current statements from the Trump administration less as "no regulation, ever," and more as "leave the oversight of US firms to American lawmakers."


What I'm reading


— The University of Amsterdam's DSA Observatory sketches out the current state of play for enforcement under the EU's online safety rules. More here.

— The United Kingdom's Ofcom regulator outlines non-binding rules for how online platforms should handle online harms against women and girls. More here.

— The White House published its so-called "Genesis Mission" to jumpstart the use of federal resources for AI-enable scientific research. More here.

— The European venture capital firm Atomico published its annual report on the state of the Continent's technology start-up technology industry. More here.



digitalpolitics.co/newsletter0…



Shai Hulud 2.0, now with a wiper flavor


In September, a new breed of malware distributed via compromised Node Package Manager (npm) packages made headlines. It was dubbed “Shai-Hulud”, and we published an in-depth analysis of it in another post. Recently, a new version was discovered.

Shai Hulud 2.0 is a type of two-stage worm-like malware that spreads by compromising npm tokens to republish trusted packages with a malicious payload. More than 800 npm packages have been infected by this version of the worm.

According to our telemetry, the victims of this campaign include individuals and organizations worldwide, with most infections observed in Russia, India, Vietnam, Brazil, China, Türkiye, and France.

Technical analysis


When a developer installs an infected npm package, the setup_bun.js script runs during the preinstall stage, as specified in the modified package.json file.

Bootstrap script


The initial-stage script setup_bun.js is left intentionally unobfuscated and well documented to masquerade as a harmless tool for installing the legitimate Bun JavaScript runtime. It checks common installation paths for Bun and, if the runtime is missing, installs it from an official source in a platform-specific manner. This seemingly routine behavior conceals its true purpose: preparing the execution environment for later stages of the malware.


The installed Bun runtime then executes the second-stage payload, bun_environment.js, a 10MB malware script obfuscated with an obfuscate.io-like tool. This script is responsible for the main malicious activity.


Stealing credentials


Shai Hulud 2.0 is built to harvest secrets from various environments. Upon execution, it immediately searches several sources for sensitive data, such as:

  • GitHub secrets: the malware searches environment variables and the GitHub CLI configuration for values starting with ghp_ or gho_. It also creates a malicious workflow yml in victim repositories, which is then used to obtain GitHub Actions secrets.
  • Cloud credentials: the malware searches for cloud credentials across AWS, Azure, and Google Cloud by querying cloud instance metadata services and using official SDKs to enumerate credentials from environment variables and local configuration files.
  • Local files: it downloads and runs the TruffleHog tool to aggressively scan the entire filesystem for credentials.

Then all the exfiltrated data is sent through the established communication channel, which we describe in more detail in the next section.


Data exfiltration through GitHub


To exfiltrate the stolen data, the malware sets up a communication channel via a public GitHub repository. For this purpose, it uses the victim’s GitHub access token if found in environment variables and the GitHub CLI configuration.


After that, the malware creates a repository with a randomly generated 18-character name and a marker in its description. This repository then serves as a data storage to which all stolen credentials and system information are uploaded.

If the token is not found, the script attempts to obtain a previously stolen token from another victim by searching through GitHub repositories for those containing the text, “Sha1-Hulud: The Second Coming.” in the description.


Worm spreading across packages


For subsequent self-replication via embedding into npm packages, the script scans .npmrc configuration files in the home directory and the current directory in an attempt to find an npm registry authorization token.

If this is successful, it validates the token by sending a probe request to the npm /-/whoami API endpoint, after which the script retrieves a list of up to 100 packages maintained by the victim.

For each package, it injects the malicious files setup_bun.js and bun_environment.js via bundleAssets and updates the package configuration by setting setup_bun.js as a pre-installation script and incrementing the package version. The modified package is then published to the npm registry.


Destructive responses to failure


If the malware fails to obtain a valid npm token and is also unable to get a valid GitHub token, making data exfiltration impossible, it triggers a destructive payload that wipes user files, primarily those in the home directory.


Our solutions detect the family described here as HEUR:Worm.Script.Shulud.gen.


Since September of this year, Kaspersky has blocked over 1700 Shai Hulud 2.0 attacks on user machines. Of these, 18.5% affected users in Russia, 10.7% occurred in India, and 9.7% in Brazil.

TOP 10 countries and territories affected by Shai Hulud 2.0 attacks (download)
We continue tracking this malicious activity and provide up-to-date information to our customers via the Kaspersky Open Source Software Threats Data Feed. The feed includes all packages affected by Shai-Hulud, as well as information on other open-source components that exhibit malicious behaviour, contain backdoors, or include undeclared capabilities.


securelist.com/shai-hulud-2-0/…



Exploits and vulnerabilities in Q3 2025


In the third quarter, attackers continued to exploit security flaws in WinRAR, while the total number of registered vulnerabilities grew again. In this report, we examine statistics on published vulnerabilities and exploits, the most common security issues impacting Windows and Linux, and the vulnerabilities being leveraged in APT attacks that lead to the launch of widespread C2 frameworks. The report utilizes anonymized Kaspersky Security Network data, which was consensually provided by our users, as well as information from open sources.

Statistics on registered vulnerabilities


This section contains statistics on registered vulnerabilities. The data is taken from cve.org.

Let us consider the number of registered CVEs by month for the last five years up to and including the third quarter of 2025.

Total published vulnerabilities by month from 2021 through 2025 (download)

As can be seen from the chart, the monthly number of vulnerabilities published in the third quarter of 2025 remains above the figures recorded in previous years. The three-month total saw over 1000 more published vulnerabilities year over year. The end of the quarter sets a rising trend in the number of registered CVEs, and we anticipate this growth to continue into the fourth quarter. Still, the overall number of published vulnerabilities is likely to drop slightly relative to the September figure by year-end

A look at the monthly distribution of vulnerabilities rated as critical upon registration (CVSS > 8.9) suggests that this metric was marginally lower in the third quarter than the 2024 figure.

Total number of critical vulnerabilities published each month from 2021 to 2025 (download)

Exploitation statistics


This section contains exploitation statistics for Q3 2025. The data draws on open sources and our telemetry.

Windows and Linux vulnerability exploitation


In Q3 2025, as before, the most common exploits targeted vulnerable Microsoft Office products.

Most Windows exploits detected by Kaspersky solutions targeted the following vulnerabilities:

  • CVE-2018-0802: a remote code execution vulnerability in the Equation Editor component
  • CVE-2017-11882: another remote code execution vulnerability, also affecting Equation Editor
  • CVE-2017-0199: a vulnerability in Microsoft Office and WordPad that allows an attacker to assume control of the system

These vulnerabilities historically have been exploited by threat actors more frequently than others, as discussed in previous reports. In the third quarter, we also observed threat actors actively exploiting Directory Traversal vulnerabilities that arise during archive unpacking in WinRAR. While the originally published exploits for these vulnerabilities are not applicable in the wild, attackers have adapted them for their needs.

  • CVE-2023-38831: a vulnerability in WinRAR that involves improper handling of objects within archive contents We discussed this vulnerability in detail in a 2024 report.
  • CVE-2025-6218 (ZDI-CAN-27198): a vulnerability that enables an attacker to specify a relative path and extract files into an arbitrary directory. A malicious actor can extract the archive into a system application or startup directory to execute malicious code. For a more detailed analysis of the vulnerability, see our Q2 2025 report.
  • CVE-2025-8088: a zero-day vulnerability similar to CVE-2025-6128, discovered during an analysis of APT attacks The attackers used NTFS Streams to circumvent controls on the directory into which files were unpacked. We will take a closer look at this vulnerability below.

It should be pointed out that vulnerabilities discovered in 2025 are rapidly catching up in popularity to those found in 2023.

All the CVEs mentioned can be exploited to gain initial access to vulnerable systems. We recommend promptly installing updates for the relevant software.

Dynamics of the number of Windows users encountering exploits, Q1 2023 — Q3 2025. The number of users who encountered exploits in Q1 2023 is taken as 100% (download)

According to our telemetry, the number of Windows users who encountered exploits increased in the third quarter compared to the previous reporting period. However, this figure is lower than that of Q3 2024.

For Linux devices, exploits for the following OS kernel vulnerabilities were detected most frequently:

  • CVE-2022-0847, also known as Dirty Pipe: a vulnerability that allows privilege escalation and enables attackers to take control of running applications
  • CVE-2019-13272: a vulnerability caused by improper handling of privilege inheritance, which can be exploited to achieve privilege escalation
  • CVE-2021-22555: a heap overflow vulnerability in the Netfilter kernel subsystem. The widespread exploitation of this vulnerability is due to its use of popular memory modification techniques: manipulating “msg_msg” primitives, which leads to a Use-After-Free security flaw.


Dynamics of the number of Linux users encountering exploits, Q1 2023 — Q3 2025. The number of users who encountered exploits in Q1 2023 is taken as 100% (download)

A look at the number of users who encountered exploits suggests that it continues to grow, and in Q3 2025, it already exceeds the Q1 2023 figure by more than six times.

It is critically important to install security patches for the Linux operating system, as it is attracting more and more attention from threat actors each year – primarily due to the growing number of user devices running Linux.

Most common published exploits


In Q3 2025, exploits targeting operating system vulnerabilities continue to predominate over those targeting other software types that we track as part of our monitoring of public research, news, and PoCs. That said, the share of browser exploits significantly increased in the third quarter, matching the share of exploits in other software not part of the operating system.

Distribution of published exploits by platform, Q1 2025 (download)

Distribution of published exploits by platform, Q2 2025 (download)

Distribution of published exploits by platform, Q3 2025 (download)

It is noteworthy that no new public exploits for Microsoft Office products appeared in Q3 2025, just as none did in Q2. However, PoCs for vulnerabilities in Microsoft SharePoint were disclosed. Since these same vulnerabilities also affect OS components, we categorized them under operating system vulnerabilities.

Vulnerability exploitation in APT attacks


We analyzed data on vulnerabilities that were exploited in APT attacks during Q3 2025. The following rankings draw on our telemetry, research, and open-source data.

TOP 10 vulnerabilities exploited in APT attacks, Q3 2025 (download)

APT attacks in Q3 2025 were dominated by zero-day vulnerabilities, which were uncovered during investigations of isolated incidents. A large wave of exploitation followed their public disclosure. Judging by the list of software containing these vulnerabilities, we are witnessing the emergence of a new go-to toolkit for gaining initial access into infrastructure and executing code both on edge devices and within operating systems. It bears mentioning that long-standing vulnerabilities, such as CVE-2017-11882, allow for the use of various data formats and exploit obfuscation to bypass detection. By contrast, most new vulnerabilities require a specific input data format, which facilitates exploit detection and enables more precise tracking of their use in protected infrastructures. Nevertheless, the risk of exploitation remains quite high, so we strongly recommend applying updates already released by vendors.

C2 frameworks


In this section, we will look at the most popular C2 frameworks used by threat actors and analyze the vulnerabilities whose exploits interacted with C2 agents in APT attacks.

The chart below shows the frequency of known C2 framework usage in attacks on users during the third quarter of 2025, according to open sources.

Top 10 C2 frameworks used by APT groups to compromise user systems in Q3 2025 (download)

Metasploit, whose share increased compared to Q2, tops the list of the most prevalent C2 frameworks from the past quarter. It is followed by Sliver and Mythic. The Empire framework also reappeared on the list after being inactive in the previous reporting period. What stands out is that Adaptix C2, although fairly new, was almost immediately embraced by attackers in real-world scenarios. Analyzed sources and samples of malicious C2 agents revealed that the following vulnerabilities were used to launch them and subsequently move within the victim’s network:

  • CVE-2020-1472, also known as ZeroLogon, allows for compromising a vulnerable operating system and executing commands as a privileged user.
  • CVE-2021-34527, also known as PrintNightmare, exploits flaws in the Windows print spooler subsystem, also enabling remote access to a vulnerable OS and high-privilege command execution.
  • CVE-2025-6218 or CVE-2025-8088 are similar Directory Traversal vulnerabilities that allow extracting files from an archive to a predefined path without the archiving utility notifying the user. The first was discovered by researchers but subsequently weaponized by attackers. The second is a zero-day vulnerability.


Interesting vulnerabilities


This section highlights the most noteworthy vulnerabilities that were publicly disclosed in Q3 2025 and have a publicly available description.

ToolShell (CVE-2025-49704 and CVE-2025-49706, CVE-2025-53770 and CVE-2025-53771): insecure deserialization and an authentication bypass


ToolShell refers to a set of vulnerabilities in Microsoft SharePoint that allow attackers to bypass authentication and gain full control over the server.

  • CVE-2025-49704 involves insecure deserialization of untrusted data, enabling attackers to execute malicious code on a vulnerable server.
  • CVE-2025-49706 allows access to the server by bypassing authentication.
  • CVE-2025-53770 is a patch bypass for CVE-2025-49704.
  • CVE-2025-53771 is a patch bypass for CVE-2025-49706.

These vulnerabilities form one of threat actors’ combinations of choice, as they allow for compromising accessible SharePoint servers with just a few requests. Importantly, they were all patched back in July, which further underscores the importance of promptly installing critical patches. A detailed description of the ToolShell vulnerabilities can be found in our blog.

CVE-2025-8088: a directory traversal vulnerability in WinRAR


CVE-2025-8088 is very similar to CVE-2025-6218, which we discussed in our previous report. In both cases, attackers use relative paths to trick WinRAR into extracting archive contents into system directories. This version of the vulnerability differs only in that the attacker exploits Alternate Data Streams (ADS) and can use environment variables in the extraction path.

CVE-2025-41244: a privilege escalation vulnerability in VMware Aria Operations and VMware Tools


Details about this vulnerability were presented by researchers who claim it was used in real-world attacks in 2024.

At the core of the vulnerability lies the fact that an attacker can substitute the command used to launch the Service Discovery component of the VMware Aria tooling or the VMware Tools utility suite. This leads to the unprivileged attacker gaining unlimited privileges on the virtual machine. The vulnerability stems from an incorrect regular expression within the get-versions.sh script in the Service Discovery component, which is responsible for identifying the service version and runs every time a new command is passed.

Conclusion and advice


The number of recorded vulnerabilities continued to rise in Q3 2025, with some being almost immediately weaponized by attackers. The trend is likely to continue in the future.

The most common exploits for Windows are primarily used for initial system access. Furthermore, it is at this stage that APT groups are actively exploiting new vulnerabilities. To hinder attackers’ access to infrastructure, organizations should regularly audit systems for vulnerabilities and apply patches in a timely manner. These measures can be simplified and automated with Kaspersky Systems Management. Kaspersky Symphony can provide comprehensive and flexible protection against cyberattacks of any complexity.


securelist.com/vulnerabilities…



Building a Microscope without Lenses


A mirrorless camera is mounted on a stand, facing downwards toward a rotating microscope stage made of wood. A pair of wires come down from the stage, and a man's hand is pointing to the stage.

It’s relatively easy to understand how optical microscopes work at low magnifications: one lens magnifies an image, the next magnifies the already-magnified image, and so on until it reaches the eye or sensor. At high magnifications, however, that model starts to fail when the feature size of the specimen nears the optical system’s diffraction limit. In a recent video, [xoreaxeax] built a simple microscope, then designed another microscope to overcome the diffraction limit without lenses or mirrors (the video is in German, but with automatic English subtitles).

The first part of the video goes over how lenses work and how they can be combined to magnify images. The first microscope was made out of camera lenses, and could resolve onion cells. The shorter the focal length of the objective lens, the stronger the magnification is, and a spherical lens gives the shortest focal length. [xoreaxeax] therefore made one by melting a bit of soda-lime glass with a torch. The picture it gave was indistinct, but highly magnified.
A roughly rectangular red pattern is shown, with brighter streaks converging toward the center.A cross section of the diffraction pattern of a laser diode shining through a pinhole, built up from images at different focal distances.
Besides the dodgy lens quality given by melting a shard of glass, at such high magnification some of the indistinctness was caused by the specimen acting as a diffraction grating and directing some light away from the objective lens. [xoreaxeax] visualized this by taking a series of pictures of a laser shining through a pinhole at different focal lengths, thus getting cross sections of the light field emanating from the pinhole. When repeating the procedure with a section of onion skin, it became apparent that diffraction was strongly scattering the light, which meant that some light was being diffracted out of the lens’s field of view, causing detail to be lost.

To recover the lost details, [xoreaxeax] eliminated the lenses and simply captured the interference pattern produced by passing light through the sample, then wrote a ptychography algorithm to reconstruct the original structure from the interference pattern. This required many images of the subject under different lighting conditions, which a rotating illumination stage provided. The algorithm was eventually able to recover a sort of image of the onion cells, but it was less than distinct. The fact that the lens-free setup was able to produce any image at all is nonetheless impressive.

To see another approach to ptychography, check out [Ben Krasnow’s] approach to increasing microscope resolution. With an electron microscope, ptychography can even image individual atoms.

youtube.com/embed/lhJhRuQsiMU?…


hackaday.com/2025/12/04/buildi…



Preventing a Mess with the Weller WDC Solder Containment Pocket


Resetting the paraffin trap. (Credit: MisterHW)Resetting the paraffin trap. (Credit: MisterHW)

Have you ever tipped all the stray bits of solder out of your tip cleaner by mistake? [MisterHW] is here with a bit pf paraffin wax to save the day.

Hand soldering can be a messy business, especially when you wipe the soldering iron tip on those common brass wool bundles that have largely come to replace moist sponges. The Weller Dry Cleaner (WDC) is one of such holders for brass wool, but the large tray in front of the opening with the brass wool has confused many as to its exact purposes. In short, it’s there so that you can slap the iron against the side to flick contaminants and excess solder off the tip.

Along with catching some of the bits of mostly solder that fly off during cleaning in the brass wool section, quite a lot of debris can be collected this way. Yet as many can attest to, it’s quite easy to flip over brass wool holders and have these bits go flying everywhere.

The trap in action. (Credit: MisterHW)The trap in action. (Credit: MisterHW)

That’s where [MisterHW]’s pit of particulate holding comes into play, using folded sheet metal and some wax (e.g. paraffin) to create a trap that serves to catch any debris that enters it and smother it in the wax. To reset the trap, simply heat it up with e.g. the iron and you’ll regain a nice fresh surface to capture the next batch of crud.

As the wax is cold when in use, even if you were to tip the holder over, it should not go careening all over your ESD-safe work surface and any parts on it, and the wax can be filtered if needed to remove the particulates. When using leaded solder alloys, this setup also helps to prevent lead-contamination of the area and generally eases clean-up as bumping or tipping a soldering iron stand no longer means weeks, months or years of accumulations scooting off everywhere.


hackaday.com/2025/12/04/preven…



Build A Pocket-Sized Wi-Fi Analyzer


Wi-Fi! It’s everywhere, and yet you can’t really see it, by virtue of the technology relying on the transmission of electromagnetic waves outside the visual spectrum. Never mind, though, because you can always build yourself a Wi-Fi analyzer to get some insight into your radio surroundings, as demonstrated by [moononournation].

The core of the build is the ESP32-C5. The popular microcontroller is well-equipped for this task with its onboard dual-band Wi-Fi hardware, even if the stock antenna on most devboards is a little underwhelming. [moononournation] has paired this with a small rectangular LCD screen running the ILI9341 controller. The graphical interface is drawn with the aid of the Arduino_GFX library. It shows a graph of access points detected in the immediate area, as well as which channels they’re using and their apparent signal strength.

If you’re just trying to get a basic read on the Wi-Fi environment in a given locale, a tool like this can prove pretty useful. If your desires are more advanced, you might leap up to tinkering in the world of software defined radio. Video after the break.

youtube.com/embed/t9VukUucfEA?…


hackaday.com/2025/12/04/build-…



Raising a GM EV1 from the Dead


Probably the biggest story in the world of old cars over the past couple of weeks has been the surfacing of a GM EV1 electric car for sale from an auto salvage yard. This was the famous electric car produced in small numbers by the automaker in the 1990s, then only made available for lease before being recalled. The vast majority were controversially crushed with a few units being donated to museums and universities in a non-functional state.

Finding an old car isn’t really a Hackaday story in itself, but now it’s landed in [The Questionable Garage]. It’s being subjected to a teardown as a prelude to its restoration, offering a unique opportunity to look at the state of the art in 1990s electric automotive technology.

The special thing about this car is that by a murky chain of events it ended up as an abandoned vehicle. GM’s legal net covers the rest of the surviving cars, but buying this car as an abandoned vehicle gives the owner legal title over it and frees him from their restrictions. The video is long, but well worth a watch as we see pieces of automotive tech never before shown in public. As we understand it the intention is to bring it to life using parts from GM’s contemporary S10 electric pickup truck — itself a rare vehicle — so we learn quite a bit about those machines too.

Along the way they find an EV1 charger hiding among a stock of pickup chargers, take us through the vehicle electronics, and find some galvanic corrosion in the car’s structure due to water ingress. The windscreen has a huge hole, which they cover with a plastic wrap in order to 3D scan so they can create a replacement.

This car will undoubtedly become a star of the automotive show circuit due to its unique status, so there will be plenty of chances to look at it from the outside in future. Seeing it this close up in parts though is as unique an opportunity as the car itself. We’ve certainly seen far more crusty conventional cars restored to the road, but without the challenge of zero parts availability and no donor cars. Keep an eye out as they bring it closer to the road.

youtube.com/embed/Xn2MJqPOmSI?…


hackaday.com/2025/12/04/raisin…



Keebin’ with Kristina: the One with the Pretty Protoypes


Illustrated Kristina with an IBM Model M keyboard floating between her hands.

Some like it flat, and there’s nothing wrong with that. What you are looking at is the first prototype of Atlas by [AsicResistor], which is still a work in progress. [AsicResistor] found the Totem to be a bit cramped, so naturally, it was time to design a keyboard from the ground up.

Image by [AsicResistor] via redditThe case is wood, if that’s not immediately obvious. This fact is easily detectable in the lovely render, but I didn’t want to show you that here.

This travel-friendly keyboard has 34 keys and dual trackpoints, one on each half. If the nubbin isn’t your thing, there’s an optional, oversized trackball, which I would totally opt for. But I would need an 8-ball instead, simply because that’s my number.

A build video is coming at some point, so watch the GitHub, I suppose, or haunt r/ergomechkeyboards.

Flat as it may be, I would totally at least give this keyboard a fair chance. There’s just something about those keycaps, for starters. (Isn’t it always the keycaps with me?) For another, I dig the pinky stagger. I’m not sure that two on each side is nearly enough thumb keys for me, however.

The Foot Roller Scroller Is Not a Crock


Sitting at a keyboard all day isn’t great for anyone, but adding in some leg and/or foot movement throughout the day is a good step in the right direction. Don’t want to just ride a bike all day under your desk? Add something useful like foot pedals.

Image by [a__b] via redditThe Kinesis Savant pedals are a set of three foot switches that are great for macros, or just pressing Shift all the time. Trust me. But [a__b] wasn’t satisfied with mere clicking, and converted their old pedals into a Bluetooth 5.0 keyboard with a big, fat scroll wheel.

Brain-wise, it has a wireless macro keyboard and an encoder from Ali, but [a__b] plans to upgrade it to a nice!nano in order to integrate it with a Glove80.

Although shown with a NautiCroc, [a__b] says the wheel works well with socks on, or bare feet. (Take it from me, the footfeel of pedals is much more accurate with no shoes on.) Interestingly, much of the inspiration was taken from sewing machines.

As of this writing, [a__b] has mapped all keys using BetterTouchTool for app-specific action, and is out there happily scrolling through pages, controlling the volume, and navigating YouTube videos. Links to CAD and STLs are coming soon.

The Centerfold: LEGO My Ergo


Image by [Flat-Razzmatazz-672] via redditThis here is a Silakka 54 split keyboard with a custom LEGO case available on Thingiverse. [Flat-Razzmatazz-672] says that it isn’t perfect (could have fooled me!), but it did take a hell of a lot of work to get everything to fit right.

As you might imagine and [Flat-Razzmatazz-672] can attest, 3D printing LEGO is weird. These studs are evidently >= 5% bigger than standard studs, because if you print it as is, the LEGO won’t fit right.

Via reddit

Do you rock a sweet set of peripherals on a screamin’ desk pad? Send me a picture along with your handle and all the gory details, and you could be featured here!

Historical Clackers: the North’s was a Striking Down-striker


Although lovely to gaze upon, the North’s typewriter was a doomed attempt at creating a visible typewriter. That is, one where a person could actually see what they were typing as they typed it.

Image via The Antikey Chop

North’s achieved this feat through the use of vertical typebars arranged in a semi-circle that would strike down onto the platen from behind, making it a rear down-striker.

In order for this arrangement to work, the paper had to be loaded, coiled into one basket, and it was fed into another, hidden basket while typing. This actually allowed the typist to view two lines at a time, although the unfortunate ribbon placement obstructed the immediate character.

The story of North’s typewriter is a fairly interesting one. For starters, it was named after Colonel John Thomas North, who wasn’t really a colonel at all. In fact, North had very little to do with the typewriter beyond bankrolling it and providing a name.

North started the company by purchasing the failed English Typewriter Company, which brought along with it a couple of inventors, who would bring the North’s to fruition. The machine was made from 1892 to 1905. In 1896, North died suddenly while eating raw oysters, though the cause of death was likely heart failure. As he was a wealthy, unpopular capitalist, conspiracy theories abounded surrounding his departure.

Finally, MoErgo Released a New Travel Keyboard, the Go60


It’s true, the MoErgo Glove80 is great for travel. And admittedly, it’s kind of big, both in and out of its (very nice) custom zipper case. But you asked, and MoErgo listened. And soon enough, there will be a new option for even sleeker travel, the Go60. Check out the full spec sheet.

Image by MoErgo via reddit

You may have noticed that it’s much flatter than the Glove80, which mimics the key wells of a Kinesis Advantage quite nicely.

Don’t worry, there are removable palm rests that are a lot like the Glove80 rests. And it doesn’t have to be flat –there is 6-step magnetic tenting (6.2° – 17°), which snaps on or off in seconds. The palm rests have 7-step tenting (6°-21.5°), and they come right off, too.

Let’s talk about those trackpads. They are Cirque 40 mm Glidepoints. They aren’t multi-touch, but they are fully integrated into ZMK and thus are fully programmable, so do what you will.

Are you as concerned about battery life as I am? It’s okay — the Go60 goes fully wired with a TRRS cable between the halves, and a USB connection from the left half to the host. Although ZMK did not support this feature, MoErgo sponsored the founder, [Pete], to develop it, and now it’s just a feature of ZMK. You’re welcome.

Interested? The Go60 will be on Kickstarter first, and then it’ll be available on the MoErgo site. Pricing hasn’t quite been worked out yet, so stay tuned on that front.

Via reddit


Got a hot tip that has like, anything to do with keyboards? Help me out by sending in a link or two. Don’t want all the Hackaday scribes to see it? Feel free to email me directly.


hackaday.com/2025/12/04/keebin…



An Introduction to Analog Filtering


One of the major difficulties in studying electricity, especially when compared to many other physical phenomena, is that it cannot be observed directly by human senses. We can manipulate it to perform various tasks and see its effects indirectly, like the ionized channels formed during lightning strikes or the resistive heating of objects, but its underlying behavior is largely hidden from view. Even mathematical descriptions can quickly become complex and counter-intuitive, obscured behind layers of math and theory. Still, [lcamtuf] has made some strides in demystifying aspects of electricity in this introduction to analog filters.

The discussion on analog filters looks at a few straightforward examples first. Starting with an resistor-capacitor (RC) filter, [lcamtuf] explains it by breaking its behavior down into steps of how the circuit behaves over time. Starting with a DC source and no load, and then removing the resistor to show just the behavior of a capacitor, shows the basics of this circuit from various perspectives. From there it moves into how it behaves when exposed to a sine wave instead of a DC source, which is key to understanding its behavior in arbitrary analog environments such as those involved in audio applications.

There’s some math underlying all of these explanations, of course, but it’s not overwhelming like a third-year electrical engineering course might be. For anyone looking to get into signal processing or even just building a really nice set of speakers for their home theater, this is an excellent primer. We’ve seen some other demonstrations of filtering data as well, like this one which demonstrates basic filtering using a microcontroller.


hackaday.com/2025/12/04/an-int…



Ore Formation: A Surface Level Look


The past few months, we’ve been giving you a quick rundown of the various ways ores form underground; now the time has come to bring that surface-level understanding to surface-level processes.

Strictly speaking, we’ve already seen one: sulfide melt deposits are associated with flood basalts and meteorite impacts, which absolutely are happening on-surface. They’re totally an igneous process, though, and so were presented in the article on magmatic ore processes.

For the most part, you can think of the various hydrothermal ore formation processes as being metamorphic in nature. That is, the fluids are causing alteration to existing rock formations; this is especially true of skarns.

There’s a third leg to that rock tripod, though: igneous, metamorphic, and sedimentary. Are there sedimentary rocks that happen to be ores? You betcha! In fact, one sedimentary process holds the most valuable ores on Earth– and as usual, it’s not likely to be restricted to this planet alone.

Placer? I hardly know ‘er!


We’re talking about placer deposits, which means we’re talking about gold. In dollar value, gold’s great expense means that these deposits are amongst the most valuable on Earth– and nearly half of the world’s gold has come out of just one of them. Gold isn’t the only mineral that can be concentrated in placer deposits, to be clear; it’s just the one everyone cares about these days, because, well, have you seen the spot price lately?

The spot price of gold going back 30 years. Oof.Oof. Data from Goldprice.org

Since we’re talking about sediments, as you might guess, this is a secondary process: the gold has to already be emplaced by one of the hydrothermal ore processes. Then the usual erosion happens: wind and water breaks down the rock, and gold gets swept downhill along with all the other little bits of rock on their way to becoming sediments. Gold, however, is much denser than silicate rocks. That’s the key here: any denser material is naturally going to be sorted out in a flow of grains. To be specific, empirical data shows that anything denser than 2.87 g/cm3 can be concentrated in a placer deposit. That would qualify a lot of the sulfide minerals the hydrothermal processes like to throw up, but unfortunately sulfides tend to be both too soft and too chemically unstable to hold up to the weathering to form placer deposits, at least on Earth since cyanobacteria polluted the atmosphere with O2.

Windswept dunes on Mars as pictured by MSL.Dry? Check. Windswept? Check. Aeolian placer deposits? Maybe!
Image: “MSL Sunset Dunes Mosaic“, NASA/JPL and Olivier de Goursac

One form of erosion is from wind, which tends to be important in dry regions – particularly the deserts of Australia and the Western USA. Wind erosion can also create placer deposits, which get called “aeolian placers”. The mechanism is fairly straightforward: lighter grains of sand are going to blow further, concentrating the heavy stuff on one side of a dune or closer to the original source rock. Given the annual global dust storms, aeolian placers may come up quite often on Mars, but the thin atmosphere might make this process less likely than you’d think.

We’ve also seen rockslides on Mars, and material moving in this matter is subject to the same physics. In a flow of grains, you’re going to have buoyancy and the heavy stuff is going to fall to the bottom and stop sooner. If the lighter material is further carried away by wind or water, we call the resulting pile of useful, heavy rock an effluvial placer deposit.

Still, on this planet at least it’s usually water doing the moving of sediments, and it’s water that’s doing the sortition. Heavy grains fall out of suspension in water more easily. This tends to happen wherever flow is disrupted: at the base of a waterfall, at a river bend, or where a river empties into a lake or the ocean. Any old Klondike or California prospector would know that that’s where you’re going to go panning for gold, but you probably wouldn’t catch a 49er calling it an “Alluvial placer deposit”. Panning itself is using the exact same physics– that’s why it, along with the fancy modern sluices people use with powered pumps, are called “placer mining”. Mars’s dry river beds may be replete with alluvial placers; so might the deltas on Titan, though on a world where water is part of the bedrock, the cryo-mineralogy would be very unfamiliar to Earthly geologists.

Back here on earth, wave action, with the repeated reversal of flow, is great at sorting grains. There aren’t any gold deposits on beaches these days because wherever they’ve been found, they were mined out very quickly. But there are many beaches where black magnetite sand has been concentrated due to its higher density to quartz. If your beach does not have magnetite, look at the grain size: even quartz grains can often get sorted by size on wavy beaches. Apparently this idea came after scientists lost their fascination with latin, as this type of deposit is referred to simply as a “beach placer” rather than a “littoral placer”.

Kondike, eat your heart out: Fifty thousand tonnes of this stuff has come out of the mines of Witwatersrand.

While we in North America might think of the Klondike or California gold rushes– both of which were sparked by placer deposits– the largest gold field in the world was actually in South Africa: the Witwatersrand Basin. Said basin is actually an ancient lake bed, Archean in origin– about three billion years old. For 260 million years or thereabouts, sediments accumulated in this lake, slowly filling it up. Those sediments were being washed out from nearby mountains that housed orogenic gold deposits. The lake bed has served to concentrate that ancient gold even further, and it’s produced a substantial fraction of the gold metal ever extracted– depending on the source, you’ll see numbers from as high as 50% to as low as 22%. Either way, that’s a lot of gold.

Witwatersrand is a bit of an anomaly; most placer deposits are much smaller than that. Indeed, that’s in part why you’ll find placer deposits only mined for truly valuable minerals like gold and gems, particularly diamonds. Sure, the process can concentrate magnetite, but it’s not usually worth the effort of stripping a beach for iron-rich sand.

The most common non-precious exception is uraninite, UO2, a uranium ore found in Archean-age placer deposits. As you might imagine, the high proportion of heavy uranium makes it a dense enough mineral to form placer deposits. I must specify Archean-age, however, because an oxygen atmosphere tends to further oxidize the uraninite into more water-soluble forms, and it gets washed to sea instead of forming deposits. On Earth, it seems there are no uraninite placers dated to after the Great Oxygenation; you wouldn’t have that problem on Mars, and the dry river beds of the red planet may well have pitchblende reserves enough for a Martian rendition of “Uranium Fever”.

If you were the Martian, would you rather find uranium or gold in those river bends?
Image: Nandes Valles valley system, ESA/DLR/FU Berlin

While uranium is produced at Witwatersrand as a byproduct of the gold mines, uranium ore can be deposited exclusively of gold. You can see that with the alluvial deposits in Canada, around Elliot Lake in Ontario, which produced millions of pounds of the uranium without a single fleck of gold, thanks to a bend in a three-billion-year-old riverbed. From a dollar-value perspective, a gold mine might be worth more, but the uranium probably did more for civilization.

Lateritization, or Why Martians Can’t Have Pop Cans


Speaking of useful for civilization, there’s another type of process acting on the surface to give us ores of less noble metals than gold. It is not mechanical, but chemical, and given that it requires hot, humid conditions with lots of water, it’s almost certainly restricted to Sol 3. As the subtitle gives it away, this process is called “lateritization” and is responsible for the only economical aluminum deposits out there, along with a significant amount of the world’s nickel reserves.

The process is fairly simple: in the hot tropics, ample rainfall will slowly leech any mobile ions out of clay soils. Ions like sodium and potassium are first to go, followed by calcium and magnesium but if the material is left on the surface long enough, and the climate stays hot and wet, chemical weathering will eventually strip away even the silica. The resulting “Laterite” rock (or clay) is rich in iron, aluminum, and sometimes nickel and/or copper. Nickel laterites are particularly prevalent in New Caledonia, where they form the basis of that island’s mining industry. Aluminum-rich laterites are called bauxite, and are the source of all Earth’s aluminum, found worldwide. More ancient laterites are likely to be found in solid form, compressed over time into sedimentary rock, but recent deposits may still have the consistency of dirt. For obvious reasons, those recent deposits tend to be preferred as cheaper to mine.

That red dirt is actually aluminum ore, from a 1980s-era operation on the island of Jamaica. Image from “Bauxite” by Paul Morris, CC BY-SA 2.0

When we talk about a “warm and wet” period in Martian history, we’re talking about the existence of liquid water on the surface of the planet– we are notably not talking about tropical conditions. Mars was likely never the kind of place you’d see lateritization, so it’s highly unlikely we will ever find bauxite on the surface of Mars. Thus future Martians will have to make due without Aluminum pop cans. Of course, iron is available in abundance there and weighs about the same as the equivalent volume of aluminum does here on Earth, so they’ll probably do just fine without it.

Most nickel has historically come from sulfide melt deposits rather than lateralization, even on Earth, so the Martians should be able to make their steel stainless. Given the ambitions some have for a certain stainless-steel rocket, that’s perhaps comforting to hear.

It’s important to emphasize, as this series comes to a close, that I’m only providing a very surface-level understanding of these surface level processes– and, indeed, of all the ore formation processes we’ve discussed in these posts. Entire monographs could be, and indeed have been written about each one. That shouldn’t be surprising, considering the depths of knowledge modern science generates. You could do an entire doctorate studying just one aspect of one of the processes we’ve talked about in this series; people have in the past, and will continue to do so for the foreseeable future. So if you’ve found these articles interesting, and are sad to see the series end– don’t worry! There’s a lot left to learn; you just have to go after it yourself.

Plus, I’m not going anywhere. At some point there are going to be more rock-related words published on this site. If you haven’t seen it before, check out Hackaday’s long-running Mining and Refining series. It’s not focused on the ores– more on what we humans do with them–but if you’ve read this far, it’s likely to appeal to you as well.


hackaday.com/2025/12/04/ore-fo…



Leroy Merlin subisce un attacco informatico: dati personali di clienti francesi compromessi


Un’episodio di cyberattacco ha interessato Leroy Merlin, coinvolgendo i dati personali di numerosi clienti in Francia, con un impatto su centinaia di migliaia di individui.

Leroy Merlin assicura che “sono state attivate misure di sicurezza supplementari” con una vigilanza rafforzata, “la protezione dei dati è una priorità assoluta per il marchio”, aggiunge la direzione, precisando che anche la CNIL (Commissione nazionale per l’informatica e le libertà) è stata informata della situazione.

I dati rubati sono relativi soprattutto a informazioni di contatto, quali date di nascita, numeri di telefono, indirizzi di posta elettronica, nomi, domicili e informazioni sul programma fedeltà.

Queste informazioni sono sufficienti a supportare campagne di phishing molto credibili, Frodi su misura e tecniche di ingegneria sociale che usano la reputazione di affidabilità di uno dei marchi più sicuri e popolari nel settore del bricolage e dei prodotti per la casa e il giardino.

L’azienda ha comunicato agli interessati la violazione riportando quanto segue: “Un attacco informatico ha recentemente preso di mira il nostro sistema informativo e alcuni dei vostri dati personali potrebbero essere trapelati all’esterno dell’azienda. Non appena l’incidente è stato rilevato, abbiamo adottato tutte le misure necessarie per impedire accessi non autorizzati e contenere la situazione. Le informazioni in questione sono i tuoi dati di contatto (nome, cognome, numero di telefono, indirizzo e-mail, indirizzo postale, data di nascita) e le informazioni relative al tuo programma fedeltà.”

Sembra che, per fortuna, le informazioni delicate come account, password e dati bancari siano state preservate. I clienti italiani sono stati risparmiati, poiché l’incidente ha coinvolto solo quelli francesi.

La CNIL ha segnalato nell’ultimo rapporto annuale che il numero di violazioni sta aumentando rapidamente nel Paese transalpino. In soli 12 mesi, più di un milione di persone sono state coinvolte e il numero di attacchi riusciti è raddoppiato, passando da 20 a 40.

L'articolo Leroy Merlin subisce un attacco informatico: dati personali di clienti francesi compromessi proviene da Red Hot Cyber.

informapirata ⁂ reshared this.



Cavi sottomarini: sempre più infrastrutture strategiche per la connessione globale


I cavi sottomarini in fibra ottica stanno diventando uno degli elementi più strategici delle relazioni internazionali. Pur essendo spesso ignorati rispetto alle tecnologie più visibili, come l’intelligenza artificiale, rappresentano l’infrastruttura attraverso cui transita la quasi totalità dei dati globali. Con oltre 1,4 milioni di chilometri di reti posate sui fondali, costituiscono la struttura essenziale che sostiene comunicazioni civili, militari e finanziarie.

Le agenzie internazionali che monitorano il settore, tra cui l’ITU e l’Agenzia europea per la sicurezza informatica, confermano che un’interruzione di questi collegamenti avrebbe ripercussioni immediate su cloud, servizi digitali, mercati e logistica. Le capacità dei satelliti non sono in grado di eguagliare la quantità di dati che i cavi riescono a trasportare, rendendoli una componente insostituibile del sistema digitale mondiale.

Di fronte a questa centralità, Stati Uniti ed Europa stanno rafforzando la protezione delle loro infrastrutture sottomarine. Washington ha discusso limitazioni alla partecipazione di Paesi considerati rivali nei progetti di posa dei cavi che collegano gli USA, mentre Bruxelles sta sviluppando nuove misure per aumentare il monitoraggio marittimo e coordinare la sicurezza con la NATO. Nessuna di queste iniziative cita minacce specifiche, ma tutte evidenziano un crescente allarme.

La vulnerabilità delle reti è emersa con chiarezza tra il 2023 e il 2024, quando due episodi – uno nel Mar Rosso e uno nel Mar Baltico– hanno causato gravi rallentamenti nelle comunicazioni intercontinentali. Indagini e rapporti della NATO hanno sottolineato come le infrastrutture europee siano esposte sia a incidenti tecnici sia al rischio di sabotaggi, in particolare nei passaggi marittimi più sensibili.

Il controllo delle rotte dei cavi è diventato un terreno di scontro tra le grandi potenze. Secondo analisi del CSIS, la competizione tra Stati Uniti e Cina si concentra soprattutto nell’Indo-Pacifico, una zona attraverso cui passa una quota significativa del traffico globale. Il costo elevato della posa e della manutenzione rende questi progetti legati a investimenti nazionali e sostegni governativi, trasformandoli in strumenti di influenza strategica.

Tra i sistemi più estesi e rilevanti c’è la serie di cavi “Sea Me We”, che collega il Sud-est asiatico all’Europa attraverso il Medio Oriente. L’evoluzione di questi progetti, insieme alla crescita di cloud computing, e-commerce e servizi digitali, ha reso indispensabile ampliare e aggiornare le reti esistenti. La necessità di manutenzione costante aumenta però i rischi, soprattutto nelle aree geopoliticamente instabili.

Parallelamente, la Cina sta ampliando la propria presenza globale attraverso i cavi sottomarini, integrandoli nella sua “Via della Seta Digitale”. Studi universitari e documenti ufficiali indicano che Pechino sta finanziando nuovi collegamenti in Africa e Asia per costruire un’infrastruttura meno dipendente dai sistemi occidentali. Questo approccio è visto da Stati Uniti ed Europa come una spinta espansionistica che potrebbe influenzare gli standard e il flusso dei dati.

Infine, anche le economie emergenti e i Paesi in via di sviluppo stanno investendo per accedere ai cavi sottomarini, considerandoli una condizione necessaria per integrarsi nell’economia digitale globale. In assenza di standard condivisi per la protezione da sabotaggi e minacce, la sicurezza resta un tema aperto.

In un mondo sempre più dipendente dai flussi di dati, il controllo e la protezione dei cavi sottomarini stanno diventando un elemento decisivo negli equilibri strategici internazionali.

L'articolo Cavi sottomarini: sempre più infrastrutture strategiche per la connessione globale proviene da Red Hot Cyber.



Le SIM anonime in Russia sono ancora possibili anche dopo l’introduzione di nuove leggi


Gli sforzi dei legislatori e delle forze dell’ordine per contrastare il riciclaggio di denaro e le procedure più complesse di verifica delle schede SIM non hanno indebolito in modo significativo la posizione dei commercianti di numeri anonimi. Questa conclusione emerge da uno studio sull’offerta nei mercati dell’elettronica di Mosca.

Nonostante l’introduzione di nuove procedure di registrazione delle schede SIM nella Federazione Russa, ottenere un numero anonimo rimane ancora semplice, inclusa la possibilità di ripristinarlo in un secondo momento. Per aggirare questi requisiti, i venditori utilizzano schede SIM aziendali registrate presso società fittizie.

Come ha scoperto Izvestia, una scheda SIM di questo tipo può essere acquistata senza particolari ostacoli.

Gli annunci di vendita si trovano solitamente sui social media o anche nei negozi stessi e nei loro siti internet. Anche altri prodotti “particolari”, come le telecamere compromesse, vengono vendute in questi market. Il prezzo per un numero “anonimo” parte dai 4.000 rubli, ovvero circa 45 euro.

Secondo le fonti della pubblicazione, i principali acquirenti di tali schede SIM sono i mercati criminali : “Il flusso incontrollato di numeri sfruttati per attività criminali proviene da cittadini stranieri. Per loro, il mercato nero non è affatto un mercato, ma una pratica comune.

I commercianti non hanno paura di vendere carte a sconosciuti e spesso operano tramite intermediari fidati.” Come ha spiegato Viktor Ievlev, responsabile della sicurezza informatica del Garda Group of Companies, le schede SIM sono intestate a società fittizie che le acquistano dagli operatori a tariffe aziendali”.

A volte vengono utilizzate connessioni tra dipendenti degli operatori o negozi di telefonia mobile. Sono possibili anche opzioni più discutibili, come registrarle a nome di persone decedute finite in database trapelati.

Tuttavia, Ievlev avverte che l’acquisto di tali carte è associato a gravi rischi, dai problemi con le forze dell’ordine e sono da ritenersi “pratiche illegali”.

L'articolo Le SIM anonime in Russia sono ancora possibili anche dopo l’introduzione di nuove leggi proviene da Red Hot Cyber.



Era ora! Microsoft corregge vulnerabilità di Windows sfruttata da 8 anni


Microsoft ha silenziosamente corretto una vulnerabilità di Windows di vecchia data, sfruttata in attacchi reali per diversi anni. L’aggiornamento è stato rilasciato nel Patch Tuesday di novembre , nonostante l’azienda fosse stata in precedenza lenta nell’affrontare il problema. Questa informazione è stata rivelata da 0patch, che ha indicato che la falla era stata sfruttata attivamente da vari gruppi dal 2017.

Il problema, denominato CVE-2025-9491, riguarda la gestione da parte di Windows delle scorciatoie LNK. Un errore dell’interfaccia utente faceva sì che parte del comando incorporato nella scorciatoia rimanesse nascosta durante la visualizzazione delle sue proprietà. Ciò consentiva l’esecuzione di codice dannoso come file innocuo. Gli esperti hanno osservato che le scorciatoie erano progettate per ingannare gli utenti, utilizzando caratteri invisibili e mascherandosi da documenti.

I primi dettagli emersero nella primavera del 2025, quando i ricercatori segnalarono che questo meccanismo veniva utilizzato da undici gruppi sponsorizzati da stati provenienti da Cina, Iran e Corea del Nord per attività di spionaggio, furto di dati e attacchi finanziari.
Paesi di origine APT che hanno sfruttato ZDI-CAN-25373 (Fonte Trendmicro)
All’epoca, la falla era nota anche come ZDI-CAN-25373. Microsoft dichiarò all’epoca che il problema non richiedeva un’attenzione immediata, citando il blocco del formato LNK in molte applicazioni Office e gli avvisi visualizzati quando si tentava di aprire tali file.

HarfangLab ha successivamente segnalato che la vulnerabilità era stata sfruttata dal gruppo XDSpy per distribuire il malware XDigo in attacchi ai governi dell’Europa orientale. Nell’autunno del 2025, Arctic Wolf ha rilevato un’altra ondata di abusi, questa volta rivolta a gruppi online cinesi che prendevano di mira istituzioni diplomatiche e governative europee e utilizzavano il malware PlugX. Microsoft ha successivamente rilasciato un chiarimento, ribadendo di non considerare il problema critico a causa della necessità di intervento da parte dell’utente e della presenza di avvisi di sistema.

Secondo 0patch, il problema andava oltre il semplice nascondere la coda del comando. Il formato di collegamento consente stringhe lunghe fino a decine di migliaia di caratteri, ma la finestra delle proprietà mostrava solo i primi 260 caratteri, troncando il resto senza preavviso. Ciò ha permesso di nascondere una parte significativa del comando eseguito. Una correzione di terze parti di 0patch ha risolto il problema in modo diverso : aggiunge un avviso quando si tenta di aprire un collegamento con argomenti più lunghi di 260 caratteri.

Un aggiornamento Microsoft ha risolto il problema espandendo il campo Destinazione in modo che venga visualizzato l’intero comando, anche se supera il limite di lunghezza precedente.

Un rappresentante dell’azienda, contattato, non ha confermato direttamente il rilascio dell’aggiornamento, ma ha fatto riferimento alle raccomandazioni generali sulla sicurezza e ha assicurato che l’azienda continua a migliorare l’interfaccia e i meccanismi di sicurezza.

L'articolo Era ora! Microsoft corregge vulnerabilità di Windows sfruttata da 8 anni proviene da Red Hot Cyber.



Una richiesta e il server è tuo! Il bug critico di React Server ha bisogno di patch immediate


Sviluppatori e amministratori di tutto il mondo stanno aggiornando urgentemente i propri server a seguito della scoperta di una vulnerabilità critica in React Server, che consente agli aggressori di eseguire codice non autenticato da remoto con una singola richiesta HTTP. L’exploit è ora disponibile al pubblico e il problema ha ricevuto il punteggio di gravità più alto, 10 su 10, sul CVSS.

React viene utilizzato attivamente sui server per accelerare il rendering di JavaScript e dei contenuti: invece di ricaricare completamente la pagina a ogni richiesta, ridisegna solo le parti modificate dell’interfaccia. Ciò consente un notevole risparmio di risorse e migliora le prestazioni dell’applicazione. Si stima che React sia utilizzato da circa il 6% di tutti i siti web e da circa il 39% degli ambienti cloud, quindi la vulnerabilità interessa un’ampia fetta dell’infrastruttura.

Gli specialisti di Wiz riferiscono che lo sfruttamento richiede una sola richiesta HTTP appositamente predisposta e i loro test hanno mostrato un tasso di successo “quasi del 100%”. Un ulteriore rischio è rappresentato dal fatto che molti framework e librerie popolari integrano React Server di default. Di conseguenza, anche le applicazioni che non utilizzano direttamente le funzionalità di React, ma il cui livello di integrazione richiama comunque codice vulnerabile, possono essere vulnerabili.

È la combinazione tra l’ampia adozione di React, la facilità di sfruttamento e il potenziale di completa acquisizione del server che ha portato al suo più alto livello di gravità. Sui social media, esperti di sicurezza e sviluppatori stanno esortando gli sviluppatori ad aggiornare senza indugio. “Di solito non lo dico, ma risolvetelo subito, accidenti “, scrive un esperto, sottolineando che la vulnerabilità di React, CVE-2025-55182, è un “10 perfetto”.

Sono interessate le versioni 19.0.1, 19.1.2 e 19.2.1 di React. Anche i componenti di terze parti che utilizzano React Server Components sono vulnerabili: i plugin Vite RSC e Parcel RSC, la versione pre-release di React Router RSC, RedwoodSDK, Waku e Next.js. La vulnerabilità per Next.js è tracciata separatamente con l’identificatore CVE-2025-66478.

Secondo Wiz e Aikido, il problema deriva dalla deserializzazione non sicura in Flight, il protocollo utilizzato nei componenti server di React. La deserializzazione è il processo di conversione di stringhe, flussi di byte e altri dati “serializzati” in oggetti e strutture in memoria. Se questo processo viene implementato in modo errato, un aggressore potrebbe iniettare dati appositamente creati che altererebbero la logica di esecuzione del codice lato server.

Gli sviluppatori di React hanno già rilasciato aggiornamenti che rafforzano la convalida dei dati in entrata e rendono più efficace il comportamento di deserializzazione per prevenire tali attacchi.

Wiz e Aikido raccomandano vivamente ad amministratori e sviluppatori di aggiornare React e tutte le dipendenze che lo utilizzano il prima possibile e di seguire attentamente le raccomandazioni dei responsabili dei framework e dei plugin sopra menzionati. Aikido consiglia inoltre di cercare informazioni sull’utilizzo di React nel codice sorgente e nei repository del progetto e di assicurarsi che tutti i componenti potenzialmente vulnerabili siano stati patchati.

L'articolo Una richiesta e il server è tuo! Il bug critico di React Server ha bisogno di patch immediate proviene da Red Hot Cyber.



CISA: Guida per l’integrazione sicura dell’AI nella tecnologia operativa (OT)


Dalla pubblicazione pubblica di ChatGPT nel novembre 2022, l’intelligenza artificiale (AI) è stata integrata in molti aspetti della società umana. Per i proprietari e gli operatori delle infrastrutture critiche, l’AI può essere utilizzata per aumentare l’efficienza e la produttività, migliorare il processo decisionale, ridurre i costi e migliorare l’esperienza dei clienti.

Nonostante i numerosi vantaggi, integrare l’AI negli ambienti di tecnologia operativa (OT) che gestiscono servizi pubblici essenziali introduce anche rischi significativi — come la deriva nel tempo dei modelli di processo OT o l’elusione dei processi di sicurezza — che i proprietari e gli operatori devono gestire con attenzione per garantire la disponibilità e l’affidabilità delle infrastrutture critiche.

L’Australian Signals Directorate, attraverso l’Australian Cyber Security Centre, insieme alla CISA e a partner internazionali e federali, ha reso pubblica una guida aggiornata in materia di sicurezza informatica, dal titolo: Principi per l’integrazione sicura dell’intelligenza artificiale nella tecnologia operativa.

L’obiettivo di questa guida è supportare gli operatori e i proprietari di infrastrutture critiche nell’integrazione sicura dell’intelligenza artificiale (IA) nei sistemi tecnologici operativi (OT), contemperando i benefici dell’IA, quali aumento dell’efficienza, ottimizzazione delle decisioni e riduzione dei costi, con le minacce specifiche che essa comporta per la sicurezza, la protezione e l’affidabilità degli ambienti OT.

Le complesse sfide in termini di sicurezza costituiscono il focus principale del documento, che prende in considerazione l’apprendimento automatico (ML), i modelli linguistici di grandi dimensioni (LLM) e gli agenti di intelligenza artificiale.

Tuttavia, le considerazioni esposte sono ugualmente pertinenti per i sistemi che si basano sulla modellazione statistica classica e sull’automazione logica.

Principi chiave per l’integrazione sicura dell’IA:


  1. Comprendere l’intelligenza artificiale: formare il personale sui rischi, gli impatti e i cicli di sviluppo sicuri dell’intelligenza artificiale.
  2. Valutare l’utilizzo dell’intelligenza artificiale in OT: valutare i casi aziendali, gestire i rischi per la sicurezza dei dati OT e affrontare le sfide di integrazione immediate e a lungo termine.
  3. Stabilire la governance dell’IA: implementare quadri di governance, testare continuamente i modelli di IA e garantire la conformità normativa.
  4. Integra sicurezza e protezione: mantieni la supervisione, garantisci la trasparenza e integra l’intelligenza artificiale nei piani di risposta agli incidenti.

I proprietari e gli operatori di infrastrutture critiche sono incoraggiati ad adottare questi principi per massimizzare i benefici dell’IA e mitigare al contempo i rischi. Per ulteriori dettagli, consultare la guida completa.

L'articolo CISA: Guida per l’integrazione sicura dell’AI nella tecnologia operativa (OT) proviene da Red Hot Cyber.



Vulnerabilità critiche nei plugin WordPress: King Addons per Elementor e Extended nel mirino


Durante il processo di registrazione, una falla di sicurezza critica (CVE-2025-8489) nel plugin King Addons per Elementor di WordPress viene sfruttata dagli aggressori, consentendo loro di acquisire privilegi amministrativi grazie ad una vulnerabilità che permette l’escalation dei privilegi.

Un componente aggiuntivo di terze parti denominato King Addons amplia le funzionalità di Elementor, un noto plugin per la creazione visiva di pagine web per siti WordPress. Con un utilizzo stimato su circa 10.000 siti web, fornisce una gamma di widget, modelli e funzionalità supplementari.

L’attività di minaccia è iniziata il 31 ottobre, appena un giorno dopo la divulgazione del problema. Finora, lo scanner di sicurezza Wordfence di Defiant, un’azienda che fornisce servizi di sicurezza per i siti web WordPress, ha bloccato oltre 48.400 tentativi di exploit.

I ricercatori hanno notato un picco nell’attività di sfruttamento tra il 9 e il 10 novembre, con due indirizzi IP più attivi: 45.61.157.120 (28.900 tentativi) e 2602:fa59:3:424::1 (16.900 tentativi).

Gli aggressori, stando alle analisi condotte da Wordfence, effettuano una richiesta contraffatta “admin-ajax.php” al fine di generare account con privilegi di amministratore non autorizzati sui siti presi di mira, specificando “user_role=administrator”.

Il difetto, identificato come CVE-2025-8489 da Peter Thaleikis, risiede nel gestore di registrazione del plugin, permettendo a qualsiasi utente registrato di assegnare a sé stesso un ruolo a propria scelta all’interno del sito web, compreso quello di amministratore, senza che siano applicate restrizioni di alcun tipo.

Si consiglia ai proprietari di siti web di eseguire l’aggiornamento alla versione 51.1.35 di King Addons, che risolve il problema CVE-2025-8489, rilasciato il 25 settembre.

Un’altra vulnerabilità critica nel plugin Extended di Advanced Custom Fields, che interessa più di 100.000 siti web WordPress, è stata segnalata dai ricercatori di Wordfence. Questa falla può consentire a un aggressore non autenticato di eseguire codice a distanza, mettendo a rischio la sicurezza dei siti interessati.

Il problema di sicurezza è stato segnalato il 18 novembre e il fornitore del plugin lo ha risolto nella versione 0.9.2 di Advanced Custom Fields: Extended, rilasciata un giorno dopo aver ricevuto il rapporto sulla vulnerabilità.

Poiché è possibile sfruttare la falla senza necessità di autenticazione attraverso una richiesta opportunamente strutturata, c’è il rischio che la diffusione pubblica di informazioni tecniche dettagliate provochi azioni dannose. Ai titolari di siti web si suggerisce di migrare alla versione più aggiornata al più presto oppure di rimuovere il plugin dai propri siti.

L'articolo Vulnerabilità critiche nei plugin WordPress: King Addons per Elementor e Extended nel mirino proviene da Red Hot Cyber.



Google testa titoli AI in Discover, ma è un disastro: confusione e disinformazione


Google sta testando titoli generati dall’intelligenza artificiale nel suo feed Discover, sostituendo i titoli delle notizie originali con quelli originali. Lo ha riferito Sean Hollister, caporedattore di The Verge, che ha notato che titoli brevi e spesso fuorvianti generati dall’intelligenza artificiale hanno iniziato ad apparire nel feed del suo smartphone al posto dei titoli dei giornali.

L’esperimento ha coinvolto il feed di notizie di Google Discover sugli smartphone Samsung Galaxy e Google Pixel. Hollister ha osservato che il sistema tenta di ridurre il significato di un post a poche parole, ma i risultati sono spesso distorti. I post su Baldur’s Gate 3 stanno ricevendo titoli che accusano i giocatori di sfruttamento minorile, mentre gli articoli sullo standard Qi2 sono impantanati in accuse di rallentamento dei vecchi Pixel.

Un articolo di Ars Technica sul prossimo prezzo della console di Valve si trasforma in un’affermazione secondo cui il prezzo sarebbe già stato rivelato, mentre un titolo accattivante sulle vendite di schede grafiche presso un rivenditore tedesco viene abbreviato in una frase sulle soluzioni AMD che presumibilmente “supereranno” Nvidia.

Compaiono semplicemente frasi prive di senso, come riferimenti a vaghi “backup” e “controversie sull’etichettatura dell’IA”, che, senza il contesto della notizia, suonano come un guazzabuglio di parole casuali.

I giornalisti sottolineano che il problema non riguarda solo la qualità della formulazione. Le pubblicazioni stanno perdendo il controllo su come i loro contenuti vengono presentati nel feed di Google e i lettori potrebbero presumere che i titoli clickbait siano creati dagli stessi redattori, poiché loghi e nomi dei media vengono ancora visualizzati accanto a essi.

La mancanza di trasparenza è una fonte particolare di frustrazione: Google contrassegna tali schede con un avviso che indica l’utilizzo di intelligenza artificiale e la possibilità di errori, ma questo avviso è visibile solo espandendo la sezione “Ulteriori dettagli”, non direttamente nel feed.

L’azienda descrive le modifiche come un piccolo esperimento di interfaccia per alcuni utenti di Discover. Secondo la portavoce di Google, Mallory DeLeon, l’obiettivo delle modifiche è rendere le informazioni su un argomento più facili da assimilare prima di arrivare sul sito. Tuttavia, la comunità giornalistica vede questo come la continuazione di una tendenza in cui il motore di ricerca e i suoi servizi correlati mantengono sempre più il pubblico all’interno dell’ecosistema Google e inviano sempre meno traffico ai siti di informazione.

In questo contesto, i team editoriali sono alla ricerca di nuove fonti di reddito, tra cui modelli di abbonamento, e avvertono che un ulteriore spostamento dell’attenzione verso l’elaborazione dei contenuti basata sull’intelligenza artificiale potrebbe solo accelerare il declino di Internet.

L'articolo Google testa titoli AI in Discover, ma è un disastro: confusione e disinformazione proviene da Red Hot Cyber.



Il meeting “Anima – Meeting del turismo religioso” si svolgerà il 6 e 7 dicembre alla Certosa di San Lorenzo a Padula, in provincia di Salerno, proponendosi come piattaforma stabile di confronto tra istituzioni, diocesi, operatori turistici, associaz…


“Che bello, parla di amore”. Lo ha detto Leone XIV al termine della visione di alcuni estratti del monologo “Pietro un uomo nel vento”, presentato questa mattina in anteprima mondiale al Maxxi di Roma e in onda il 10 dicembre su Rai1.