dip 038, #sintassi , #davidlynch slowforward.net/2025/01/22/dip…

slowforward

2025-01-22 18:13:14

dip 038, sintassi

questa annotazione può sembrar cadere qui out of the blue (e forse un po’ è così), ma va detto – o sono persuaso possa essere detto – che:

c’è un modo specifico di sentire, di avvertire la sintassi, e di naturalmente tornirne i labirinti, che è in profondità analogo al lavoro di Lynch non soltanto con la macchina da presa e con determinate sue carrellate lentissime in climax o anticlimax (per esempio), ma proprio con la gestione della trama, intesa come:

tessuto che non solo si smaglia ma si riannoda in punti imprevedibili, come un abito non euclideo. o non sempre – non tutto – euclideo.

a differenza della radice, verticale e gerarchica, il rizoma è intersecante, trasverso, anarchico e orizzontale.

in questo, una certa modalità della ricerca letteraria, che soprattutto metto al lavoro con prose brevi in un libro che uscirà prima dell’estate, le asimmetrie e astrazioni (e torsioni) sintattiche che ho sperimentato dialogano, credo proficuamente, con una idea post-novecentesca di montaggio, frammentazione e ripresa di unità.

vorrò/vorrei poi sempre più che una quota forte, alta, di indeterminazione connotasse i materiali dei prossimi testi. (ma “textus” è un vocabolo inesatto, e la parentesi resta aperta

#111 #DavidLynch #dip #dip038 #dip038 #Lynch #sintassi

Those who indulge in trading card games know that building the best deck is the key to victory. What exactly that entails is a mystery to us muggles, but keeping track of your cards is a vital part of the process, one that this DIY card scanner (original German; English translation) seeks to automate.

At its heart, [Fraens]’ card scanner is all about paper handling, which is always an engineering task fraught with peril. Cards like those for Magic: The Gathering and other TCGs are meant to be handled by human hands, and automating the task of flipping through them presents some challenges. [Fraens] uses a pair of motorized 3D-printed rollers with O-rings to form a conveyor belt that can pull one card at a time off the bottom of a deck. An adjustable retaining roller made from the most adorable linear bearing we’ve ever seen ensures that only one card at a time is pulled from the hopper onto an imaging platen. An adjustable mount holds a smartphone to take a picture of the card, which is fed into an app that extracts all the details and categorizes the cards in the deck.

Aside from the card handling mechanism, there are some pretty slick details to this build. The first is that [Fraens] noticed that the glossy finish on some cards interfered with scanning, leading him to add a diffused LED ringlight to the rig. If an image isn’t scannable, the light goes through a process of dimming and switching colors until a good scan is achieved. Also, to avoid the need to modify the existing TCG deck management app, [Fraens] added a microphone to the control side of the scanner that listens for the sounds the app makes when it scans cards. And if Magic isn’t your thing, the basic mechanism could easily be modified to scan everything from business cards to old family photos.

youtube.com/embed/dl2RyKrg4pI?…

hackaday.com/2025/03/26/3d-pri…

This week, Jonathan Bennett chats with Neal Gompa about Fedora 42 and KDE! What’s new, what’s coming, and why is flagship status such a big deal?

• Website: neal.gompa.dev/
• GitHub Sponsors: github.com/sponsors/Conan-Kudo
• Neal’s business (Velocity Limitless): velocitylimitless.com/
• Neal’s podcast (Sudo Show): tuxdigital.com/sudoshow

youtube.com/embed/xwgqPwsjd0g?…

Did you know you can watch the live recording of the show right on our YouTube Channel? Have someone you’d like us to interview? Let us know, or contact the guest and have them contact us! Take a look at the schedule here.

play.libsyn.com/embed/episode/…

Direct Download in DRM-free MP3.

If you’d rather read along, here’s the transcript for this week’s episode.

Places to follow the FLOSS Weekly Podcast:

• Spotify
• RSS

Theme music: “Newer Wave” Kevin MacLeod (incompetech.com)

Licensed under Creative Commons: By Attribution 4.0 License

hackaday.com/2025/03/26/floss-…

E ora di definire nuove policy e procedure. E’ ora di regolamentare l’utilizzo dell’intelligenza artificiale generativa all’interno delle aziende.

Il Generative AI Cloud and Threat Report 2025, pubblicato da Netskope Threat Labs, ha evidenziato un aumento preoccupante nell’uso di applicazioni di intelligenza artificiale generativa (genAI) da parte degli utenti aziendali. Nell’ultimo anno, il volume di dati inviati a queste app è cresciuto di 30 volte, comprendendo informazioni altamente sensibili come codice sorgente, dati regolamentati e proprietà intellettuale. Questa tendenza amplifica il rischio di violazioni dei dati, problemi di conformità e furto di informazioni critiche.

Uno degli aspetti più critici emersi dal rapporto è l’uso diffuso di account personali per accedere alle applicazioni genAI. Il 72% degli utenti aziendali utilizza strumenti genAI per scopi lavorativi tramite account non gestiti dall’organizzazione, creando un grave punto cieco in termini di sicurezza. James Robinson, CISO di Netskope, sottolinea nel report che nonostante gli sforzi per implementare strumenti ufficiali, la cosiddetta “shadow AI” è diventata una nuova sfida per l’IT, mettendo a rischio il controllo aziendale sui dati.

L’analisi ha rivelato che 3 utenti aziendali su 4 caricano dati sensibili su applicazioni di genAI, incluse password e chiavi di accesso. Netskope ha monitorato 317 diverse piattaforme genAI, tra cui strumenti noti come ChatGPT, Google Gemini e GitHub Copilot. La crescente integrazione di queste tecnologie negli ambienti aziendali ha reso fondamentale rafforzare la governance e il controllo, per evitare perdite involontarie di dati o violazioni della sicurezza.

Un altro dato significativo riguarda il cambiamento nell’adozione di infrastrutture locali per la genAI. Il numero di organizzazioni che utilizzano genAI on-premise è salito da meno dell’1% al 54% in un solo anno. Questo riduce il rischio di esposizione a terze parti, ma introduce nuove vulnerabilità legate alla gestione interna dei dati, alle catene di fornitura e alle potenziali fughe di informazioni. I team di sicurezza devono affrontare queste sfide con strategie più sofisticate, superando approcci di semplice blocco dell’accesso.

Per mitigare i rischi, Netskope raccomanda alle aziende di rivedere e personalizzare i propri framework di sicurezza per l’intelligenza artificiale. Le misure chiave includono la valutazione dell’uso di genAI, il rafforzamento dei controlli sulle applicazioni e la gestione più sicura dell’infrastruttura locale. L’adozione di strumenti avanzati di protezione basati sull’intelligenza artificiale sarà fondamentale per garantire un utilizzo sicuro e responsabile di queste tecnologie nel contesto aziendale.

L'articolo L’AI Sa Più Password di Te! 3 Impiegati su 4 le Condividono con gli LLM proviene da il blog della sicurezza informatica.

In Europa si sta sviluppando l’interesse per un sistema operativo realizzato per le agenzie governative. Il progetto EU OS propone di creare una distribuzione Linux immutabile basata sull’ambiente desktop KDE con un’interfaccia che ricorda quella di Windows. Autore dell’idea è il Dott. Robert Riemann, dipendente dell’Autorità europea di vigilanza sulla protezione dei dati (GEPD).

In questa fase, EU OS non è un sistema operativo pronto all’uso, bensì dettagliato. documentazione. Descrive le funzionalità richieste, i metodi di distribuzione e amministrazione, nonché i principi di collaborazione con gli utenti. Il progetto è pensato per le organizzazioni di medie dimensioni, fino a diverse centinaia di dipendenti.

Nonostante l’attenzione rivolta all’Europa, si propone di prendere come base la distribuzione americana Fedora, o più precisamente la sua versione non modificabile Kinoite con la shell KDE. Questa scelta solleva interrogativi, soprattutto alla luce delle attuali tensioni geopolitiche. Sarebbe più saggio utilizzare sviluppi europei, ad esempio openSUSE. Tuttavia, Fedora Kinoite è davvero uno dei sistemi immutabili più maturi: le sue prime versioni sono apparse più di quattro anni fa.

Un sistema operativo immutabile è un OS in cui i file di base non possono essere modificati o sovrascritti durante il normale utilizzo. Questo significa che le modifiche al sistema vengono applicate solo tramite immagini predefinite, rendendo più sicuri gli aggiornamenti e riducendo il rischio di corruzione o malware. Se qualcosa va storto, è possibile ripristinare rapidamente una versione precedente senza compromettere la stabilità del sistema. Questo approccio è particolarmente utile in ambienti aziendali e governativi, dove l’affidabilità e la sicurezza sono prioritarie. Alcuni esempi di sistemi immutabili sono Fedora Silverblue/Kinoite, openSUSE MicroOS/Kalpa e ChromeOS.

I creatori hanno studiato attentamente l’esperienza di precedenti iniziative volte a migrare le agenzie governative verso Linux. Tra questi ci sono il progetto LiMux con sede a Monaco di Baviera, attivo dal 2004 al 2017, GendBuntu della gendarmeria francese e Linux Plus 1 nello stato tedesco dello Schleswig-Holstein.

Gli esperti, tuttavia, notano che alcune delle soluzioni tecniche del progetto appaiono controverse. L’ambiente desktop KDE Plasma potrebbe essere troppo complesso per un ambiente aziendale altamente regolamentato. Sebbene la versione immutabile di Fedora sia piuttosto affidabile, esistono alternative europee, come Desktop di Kalpa basato su openSUSE.

Domande più serie sorgono proprio dal concetto di un sistema operativo locale pienamente funzionale. Nell’era degli attacchi ransomware, questo approccio sembra obsoleto. La comunità del software libero dovrebbe creare un equivalente di ChromeOS: un sistema semplice, minimalista e a doppia ridondanza, in grado di funzionare con i server cloud tramite protocolli aperti. Tutti i componenti necessari esistono già, non resta che combinarli correttamente.

Vale la pena notare che le versioni open source esistenti di ChromeOS, come ChromiumOS, ChromeOS Flex o FydeOS, non risolvono questo problema. Funzionano solo con i servizi cloud di Google e non supportano standard di autenticazione aperti come LDAP o OpenID, la sincronizzazione dei file tramite WebDAV o l’archiviazione di segnalibri, password e impostazioni utente su server indipendenti

L'articolo Windows Verrai Licenziato? Il Pinguino come Sistema Operativo per le Agenzie Governative Europee proviene da il blog della sicurezza informatica.

Printed circuit boards were once so simple. One or two layers of copper etched on a rectangular fiberglass substrate, with a few holes drilled in key locations so components could be soldered into place. They were functional objects, nothing more—built only for the sake of the circuit itself.

Fast forward to today, and so much has changed. Boards sprout so many layers, often more than 10, and all kinds of fancy geometric features for purposes both practical and pretty. But what catches they eye more than that, other than rich, saturated color? [Joseph Long] came to the 2024 Hackaday Supercon to educate us on the new world of full color PCBs.

youtube.com/embed/LOSMH_EV6pQ?…

[Joseph] begins his talk with an explanation of terminology. We often look at a PCB and cite its color—say, green for example. As [Joseph] explains, the color comes from the solder mask layer—so called for its job in ensuring solder can only go where it’s supposed to go. The solder mask sits atop the copper layer, but beneath the silk screen which has all the component outlines and part labels.

Solder mask was traditionally green, and this is still the most common color you’ll find in the majority of electronics. However, in recent decades, the available gamut of colors has increased. Now, you can routinely order yellow, blue, purple, and red solder masks quite easily, as well as black or white if you’re so inclined. As some creative makers have found, when designing a board, it’s possible to get several colors into a design even if you’re just using one color of solder mask. That’s because the solder mask appears in slightly different shades when it’s laid over the bare fiberglass of the PCB, versus being laid over copper, for example. Add in white silkscreen and you’ve got quite a lot to work with.

PCB Color Palette
byu/Half_Slab_Conspiracy inPrintedCircuitBoard

Different colors are achievable on a PCB even just by using a single soldermask color.
We’re used to having a choice of color on our PCB orders today, but so much more is possible.
But what if you want more? What if you want real color? [Joseph] realized this could be possible when he found out that PCB board houses were already using inkjet-like printers to lay down silkscreen layers on small-run boards. Since there was already a printer involved in the board production process, wouldn’t it be simple to start printing on circuit boards in full color?

As it turns out, this was very practical. Two big Shenzhen board houses—JLCPCB and PCBWay—both started delivering color printed boards in 2024. The method involved using a white solder mask layer, with a full-color “silkscreen” layer printed on top using UV-cured ink. Using this ink was a particular key to unlocking full color PCBs. The UV-cured inks are more robust under the tough conditions PCBs face, such as the high temperatures during reflow or hand soldering.

Color printing PCBs might sound trivial and only relevant for cosmetic purposes, but [Joseph] points out it has lots of practical applications too. You can easily color code pinouts and traces right on the the board, a feature that has obvious engineering value. You can even use photorealistic footprints to indicate where other board-level modules should be soldered in, too, making assembly more intuitive. Plus, full color boards are fun—don’t discount that!

[Joseph] likes using the full-color prints to aid in assembly, by using far more realistic footprints for items like board-scale modules and batteries.[Joseph] is also a big fan of the SAO format, having designed several compatible boards himself. At his talk, he showed off special “extender” boards of his own creation and offered giveaways to attendees.If you’re wondering how to get started, [Joseph]’s talk covers all the important ground. He goes over the workflow for doing color PCBs with typical board houses. As the main suppliers in this area, PCBWay and JLCPCB both have slightly different ways of working with design files for color boards. Obviously, creating a color board involves making images outside of your traditional board design software. It’s straightforward enough, but you have to follow some careful practices to ensure your images are printed in the right size and right orientation to match the rest of your PCB design. PCBWay lets you make your own images and submit them with your Gerber files from whatever board design tool, while JLCPCB requires you to produce your PCB within their EasyEDA design software and put the graphics directly in there. [Joseph] also explains the costs involved for printing these boards, which does come at a premium relative to traditional boards.

As a bonus, we even get to see some of Joseph’s awesome color boards. The graphics are stunning—they really show the potential of full-color PCBs and how they can elevate a project or a fun badge design. If you’re eager to try this out, go ahead and watch [Joseph]’s primer and dive in for yourself!

hackaday.com/2025/03/26/superc…

These days we don’t get too fussed about miniaturized electronics, not when you can put an entire processor and analog circuitry on a chip the size of a grain of sand. Things were quite different back in the 1980s, with the idea of a credit card-sized radio almost preposterous. This didn’t stop the engineers over at Casio from having a go at this and many other nutty ideas, with [Matt] from Techmoan having a go at taking one of these miniaturized marvels apart.
The Casio FM Stereo radio in happier days. (Credit: Techmoan, YouTube)
On the chopping block is the FM stereo device that was featured in a previous episode. Out of the four credit card-sized radios in that video, the one with the rechargeable battery (obviously) had ceased to work, so it was the obvious choice for a teardown. This mostly meant peeling off the glued-on top and bottom, after which the circuitry became visible.

In addition to the battery with a heavily corroded contact, the thin PCB contains a grand total of three ICs in addition to the analog circuitry. These were identified by [Spritetm] as an AM/FM tuner system IC (TA7792), an FM PLL MPX (TA7766AF) and a headphone amplifier (TA7767F), all of them manufactured by Toshiba.

Although [Matt] reckons this was a destructive teardown, we’re looking forward to the repair video where a fresh cell is soldered in and the case glued back together.

youtube.com/embed/XX_wOOvByPs?…

hackaday.com/2025/03/26/teardo…

You might think you don’t need a hearing aid, and you might be right. But in general, hearing loss eventually comes to all of us. In fact, you progressively lose hearing every year, which is why kids can have high-pitched ringtones their parents can’t hear.

You’d think hearing aids would be pretty simple, right? After all, we know how to pick up sounds, amplify them, and play them back. But there’s a lot more to it. Hearing aids need to be small, comfortable, have great battery life, and cram a microphone and speaker into a small area. That also can lead to problems with feedback, which can be very uncomfortable for the user. In addition, they need to handle very soft and loud sounds and accommodate devices like telephones.

Although early hearing aids just made sound louder and, possibly, blocked unwanted sound, modern devices will try to increase volume only in certain bands where the user has hearing loss. They may also employ sophisticated methods to block or reduce noise.

A Brief History

Hearing loss is nothing new. Ear trumpets appeared around the 17th century. These were just simple sound baffles that directed sound to your ear and, perhaps, cut some noise out that wasn’t in the trumpet’s direction.

The modern hearing aid dates back to the akouphone in 1895. [Miller Hutchison] developed the device for a friend who was deaf from a bout of scarlet fever. It was bulky — sitting on a table top — and used a carbon microphone, but it did work. He was also able to sell several models to royalty, many of whom suffered from hereditary deafness. This included Denmark’s Queen Alexandra, who, reportedly, was very impressed with the results.
The Acousticon microphone (left) and complete unit (right) (From Hawkins Electrical Guide #7, 1923)
Around 1902, [Hutchison] changed the device’s name to the acousticon, making it more portable with battery power. Despite impressive marketing, not all medical professionals were sold. If you were totally deaf, the device did nothing, unsurprisingly. In addition, the bulky batteries required frequent replacement, and the frequency response was poor.

It was still better than nothing, and the invention also led to the massacon and akoulalion that converted sound into vibration for the profoundly deaf. He later sold the rights for the acousticon to [Kelley Turner], who would not only improve the device, but also use the technology to launch the dictograph, which was a well-known office machine for many years.

Modern Times

The Zenith Miniature 75 (photo by [France1978] CC-BY-SA-2.0).Amplified hearing aids appeared around 1913, but they were still large boxes. By 1920, the vactuphone used vacuum tubes to perform amplification. At “only” seven pounds, the vactuphone was considered quite portable.

Keep in mind that portable hearing aids in the 1920s was a relative term. Typically, you’d have a unit carried in a bag or hung around your neck. World War II brought advances in minaturization which benefited hearing aids like the Zenith Miniature 75.

Transistors, of course, changed everything, including hearing aids. The Sonotone 1010, which appeared in 1952, used both transistors and tubes. Early transistor units were known to fail early due to moisture and heat. Silicon transistors and encapsulation helped.

Naturally, all of these hearing aids were analog as were the earliest IC-based devices. However, with the advent of ICs, it was possible to use digital techniques.
Patent drawing from 1984 — Hardly portable!
The path to digital hearing aids was difficult. In the 1970s, large computers could program digital elements in hearing aids to tune the device to set frequency bands and gains.

By 1980, several groups were experimenting with real digital hearing aids, although many of them had wireless links to real computers. A fully digital hearing aid first appeared in a 1984 patent, but it wasn’t tiny. Since then, things have gotten smaller and more capable.

Physical Form

Hearing aids went from table-top devices, to boxes hanging on necks. Getting smaller devices allowed for small boxes that hug the back of the ear with the earpiece into the ear canal.

With even smaller devices, the entire apparatus can be placed in the ear canal. Many of these go so deeply into the ear that they are largely invisible. There are also hearing aids that can surgically attach to your skull using a titanium post embedded in the bone. This can transmit sound even to people who can’t hear sound directly since it relies on bone conduction.

Other places to find hearing aids are built into thick glasses frames. Doctors with hearing problems can opt for stethoscopes with integrated hearing aids.

Modern hearing aids sometimes have rechargeable batteries. Otherwise, there will be some kind of small battery. There was a time that mercury cells were common, but with those banned in most places, many aids now take zinc-air batteries that deliver about 1.4 V.

We hear from an 8th grader that you can make hearing aid batteries last longer by peeling the sticker from them and waiting five minutes before installing them. Apparently, giving them a little time to mix with the air helps them.

What’s Next?

On the market today are hearing aids that use neural networks, have Bluetooth connections, and use other high tech tricks. We’ve looked at the insides of a hearing aid and why they cost so much before. If you want to roll your own, there is an open source design.

hackaday.com/2025/03/26/tech-i…

Clearview AI ha ricevuto l’approvazione giudiziaria definitiva per un accordo che fornisce una quota azionaria del 23% della società a una categoria di consumatori che hanno affermato che la società di riconoscimento facciale ha utilizzato in modo improprio le immagini delle persone senza il loro consenso riporta Bloomberg.

Approvando l’accordo giovedì, la giudice Sharon Johnson Coleman della Corte distrettuale degli Stati Uniti per il distretto settentrionale dell’Illinois ha stimato che la quota dell’accordo valesse 51,75 milioni di dollari, sulla base di una valutazione di 225 milioni di dollari di gennaio 2024.

“Certo, la natura di una quota azionaria è che potrebbe ridursi o crescere a seconda delle performance della società”, ha detto Coleman nel suo ordine. Ma “Clearview è ottimista sulla potenziale crescita della società in base al mercato disponibile”, ha detto.

In assenza di un accordo, non era chiaro se Clearview avrebbe potuto pagare milioni di dollari come parte di una sentenza, o se avesse i fondi per superare il processo, secondo un giudice in pensione che ha facilitato le negoziazioni per la conciliazione.

Clearview ha raggiunto questo accordo per risolvere le cause legali consolidate secondo cui l’azienda avrebbe violato l’Illinois Biometric Information Privacy Act estraendo miliardi di foto online e inserendole in un database di riconoscimento facciale.

L’accordo prevede la nomina di un supervisore con il diritto di ispezionare le finanze di Clearview e di vendere la quota, al fine di proteggere gli interessi della class action che ha concluso l’accordo. Coleman ha ritenuto l’accordo equo, ragionevole e adeguato, nonostante le obiezioni politiche di 22 stati e del Distretto di Columbia, nonché dei gruppi di pressione.

L'articolo 225 milioni di dollari per l’Uso Illecito delle Foto Online. Clearview perde la class Action e cede il 23% proviene da il blog della sicurezza informatica.

SUPPORTED BY

THIS IS DIGITAL POLITICS. I'm Mark Scott, and continued my Euro-trash existence this week in Geneva where I'm moderated a panel on March 24 on tech sovereignty and data governance. I'll include a write-up in next week's newsletter.

Talking of events, I'll also be co-hosting a tech policy meet-up in hipster East London on March 27 at 6:30pm. There are a few spots left for this (free) event. Sign up here.

— We're living through an era of 'tech sovereignty.' No one knows what that concept means — and that's quickly turning into a problem.

— Brussels forced Apple to open up to competitors. That's going to help many US firms that, in principle, oppose the bloc's competition revamp.

— In what must be the least-shocking fact about the latest AI models, almost none of the data used to train these systems comes from Global Majority countries.

Let's get started.

Tech sovereignty in an era of zero-sum geopolitics

MAYBE IT'S BECAUSE I WAS IN SWITZERLAND to talk about this topic, but we need to focus on tech sovereignty. Bear with me. For most of us, this concept is either unknown or irrelevant. Or possibly both. But over the last five years, policymakers and lawmakers — first in Europe, but increasingly everywhere — have embraced this catch-all term for efforts by individual governments to regain control over parts of the technology industry that have historically been left to the private sector.

Think the United States (or European Union) Chips Act, or efforts to bring back high-end semiconductor manufacturing to the "homeland." Think Washington's Joe Biden-era export controls to stop Beijing getting hold of next generation chip manufacturing equipment. Think Brussels' litany of initiatives — from the creation of so-called 'data spaces' to the (badly named) AI 'gigafactories' — to give itself a seat at the global table of tech powers.

At its core, tech sovereignty is a realization by elected officials that they are no longer in control. They see complex technological global supply chains, the rise of world-spanning tech giants and the influx of billions of dollars in private capital and worry their voters (and homegrown companies) won't see the economic and social benefits of how tech has become so ingrained in everything from buying a car to sending your child to school.

Well, maybe that's one (slightly cynical) definition. After more than five years since 'tech sovereignty' became a thing, governments are still grappling with exactly what it means, how to implement it and what the consequences will be when everyone from London to Brasilia wants to "onshore" tech to boost their local interests.

Before 2025, that remained almost exclusively a headache for uber-policy types (like myself.) But this year has shown, already, that we are living in a more transactional, zero-sum mercantilist world where all elected leaders — and not just US President Donald Trump — are willing to use all the levers at their disposal to reshape the world order to their needs.

**A message from Microsoft** Each day, millions of people use generative AI. Abusive AI-generated content, however, can present risks to vulnerable groups such as women, children, and older adults. In a new white paper, developed in consultation with civil society, we present actionable policy recommendations to promote a safer digital environment.**

That means, inevitably, revisiting how we define 'tech sovereignty' because, like it or not, how we collectively approach the topic will have significant real-world implications for how technology is developed, governed and used in the years to come.

If done well, it could build upon the core tenets of what made the internet such a game-changing technology: open, rights-based core infrastructure that allowed anyone (read: with money and technical capacity) to build whatever they wanted, however they wanted.

If done poorly, it could undermine those key principles that have made technology crucial to both economic and social benefits for all.

Case in point: if a country decides to keep all of its citizens' data within national borders — a term known as data localization — for either commercial or national security reasons, then it makes it harder to trade, based on a reduction of global data flows, and starts to cut off specific countries from the now-fraying world order. This is not hypothetical: Russia, Nigeria, India and China are among states that already have such rules on the books.

What is urgently needed is an honest conversation about what people mean by 'tech sovereignty.' Currently, that falls into two camps.

Camp One leans toward isolationism. In this world, politicians funnel public cash into homegrown 'tech champions' that use siloed-off local data and technical skills to create services/products that are then sold worldwide in a race to build global giants.

Camp Two relies on each country shifting to tech-related areas where it can compete globally (eg: Taiwan/South Korea on microchips; Vietnam on device manufacturing), and then opening up each market to overseas competition. The goal isn't to own everything in tech. It's about figuring out where you can compete, globally, while giving local citizens access to (cheap) outside services/products that improve their daily lives.

You can probably figure out which version of 'tech sovereignty' would be my preference.

Before I get angry emails, I realize there's a lot of nuance that lies between those two extreme positions. Those who want to create a so-called "Euro stack," for instance, would probably argue their efforts are about giving Europe greater autonomy at a time when the US is not perceived as a trusted partner. Those in Brazil supportive of the country's data localization mandate would likely say such provisions are about keeping local's personal information safe under national laws.

I get it. Everyone has a reason why their version of "tech sovereignty" is OK, while everyone else's take is blatant protectionism.

Thanks for reading Digital Politics. If you've been forwarded this newsletter (and like what you've read), please sign up here. For those already subscribed, reach out on digitalpolitics@protonmail.com

But here's the problem with that. This ongoing nibbling at what has made technology an inherent force for good (despite, ahem, some significant downsides) has placed increased onus on equating national power as the only mechanism to get things done. That is especially true, in 2025, when long-standing allies are starting to not trust each other, and retaliatory tariffs are leading us toward a potential global trade war.

What I would prefer to see is a recognition by lawmakers about what they can — and what they can not — change when it comes to tech. Yes, much of the current global power dynamics mean the likes of the US, China and Europe have more say than other parts of the globe. That is not something, unfortunately, that will change overnight.

But while it's 100 percent legitimate for national leaders to want greater control of various forms of technology, I don't see how the ubiquitous calls to spend public money to "bring back" global supply chains to national shores as something that will achieve that.

First, it won't — given that these complex systems have grown over decades and won't just change quickly. And second, it will lead to short-term higher prices for consumers because of the inevitable cost hikes that will result from spending over the odds to onshore manufacturing when other countries can just do it cheaper (and faster.)

"Tech sovereignty" is a concept that sounds good as a talking point, but fails to deliver when confronted with reality. Yes, some form of greater control (or, at least, the semblance of control) over global tech forces is probably good for democracy, writ large. That's especially true for countries beyond the US and China that are net-takers of technology, at a global stage.

But you don't achieve that by putting up barriers to outsiders and investing public funds to develop clunky national champions that will struggle to compete worldwide.

What would be better is to set out a positive definition of 'tech sovereignty' that builds on what has worked for almost everyone over the last 80 years. Caveat: I understand that is a difficult pitch, politically, given the current geopolitical climate.

That would include: reaffirming open global markets based on right-based digital regulation that allows each country to 1) promote their own unique tech-related specialisms, both home and abroad and 2) allow national lawmakers to step in, where appropriate, when global tech forces undermine the rule of law or other key tenets within a nation state.

We already have such systems in other sectors like financial services and pharmaceuticals — and no one (at least not yet in 2025!) makes much political capital in undermining how those industries currently operate. Yes, tech is somewhat different as it's nominally not a separate industry. But, I would argue, neither is financial services.

Unfortunately, I don't see that positive agenda in any of the ongoing 'tech sovereignty' discussions that have become embedded in the geopolitical tensions of early 2025. That goes from Trump's MAGA approach to maintaining "US dominance" over AI to European Commission president Ursula von der Leyen's pitch to make the EU the hub for the next technological revolution.

That is a shame.

It's a shame because it undermines what has been built over the last 80 in so many tech-related fields that have benefited so many people worldwide. And it's a shame because it equally foretells a growing "splinternet" between countries/regions that solely focus on their short-term interests — without recognizing what damage that will produce over the mid-term.

Chart of the Week

THE LATEST ARTIFICIAL INTELLIGENCE MODELS already skew toward more developed countries. But researchers analyzed the most common datasets used to build these systems, from 1990 to 2024, to figure out where that information actually came from.

Not surprisingly, regions like Africa and South America were massively underrepresented, both on the number of datasets (see "by count" below) from those regions and the amount of information (see "by tokens or hours" below) included from those parts of the world.

That's a problem when next generation AI models are being rolled out globally in ways that won't meet regionally-specific needs because of a lack of local data baked into these complex systems.

The darker the part of the maps below, the more data was used from that region to train AI models.
Source: The Data Provenance Initiative

The complexities of antitrust enforcement

WHEN THE EUROPEAN COMMISSION ANNOUNCEDlast week it had forced Apple to make changes to comply with the bloc's new competition rules, the iPhone maker was quick to cry foul. The decision, according to the company, "wraps us in red tape, slowing down Apple's ability to innovate for users in Europe and forcing us to give away our new features for free to companies who don't have to play by the same rules."

Yet in many parts of the global tech world — including inside companies that equally dislike the EU's Digital Markets Act— there were cheers of victory. The split response highlights how these new competition rules, which allow European regulators to step into online markets before one specific company becomes too dominant, aren't as easy to define as many first thought.

First, a quick backstory. Last year, the European Commission's competition enforcers opened an investigation into how Apple allowed rival firms to interact with its products. On March 19, Brussels then ordered the iPhone maker to make it easier for non-Apple devices to connect to the company's products. It also demanded the Cupertino-based firm to provide its technical specs to outsiders so they could build services which more easily interact with Apple's operating systems.

**A message from Microsoft**New technologies like AI supercharge creativity, business, and more. At the same time, we must take steps to ensure AI is resistant to abuse. Our latest white paper, "Protecting the Public from Abusive AI-Generated Content across the EU," highlights the weaponization of women’s nonconsensual imagery, AI-powered scams and financial fraud targeting older adults, and the proliferation of synthetic child sexual abuse.

The paper outlines steps Microsoft is taking to combat these risks and provides recommendations as to how the EU's existing regulatory framework can be used to combat the abuse of AI-generated content by bad actors. We thank Women Political Leaders, the MenABLE project, the Internet Watch Foundation, the WeProtect Global Alliance, and the European Senior’s Union for their important work and support. Click here to read more.**

What does that mean? Over the next 12 months (caveat: Apple may still appeal these changes), it will become easier, say, for Garmin smartwatches to connect seamlessly with your iPhone — just as an Apple watch currently does. Rival apps will also be able to take advantage of Apple's technical wizardry to compete more directly with the company's own services that work hand-in-glove with its in-house software.

You can understand why Apple is not a fan. But, equally, it will be a boon for the likes of Meta and Alphabet, as well as scores of smaller tech firms, that have long complained that Apple creates artificial technical barriers so that rival devices/apps just don't work as well as the iPhone maker's own offerings. Mark Zuckerberg, Meta's chief executive, even called out Apple in January over how it didn't allow other headphones to connect as well as the firm's (expensive) devices.

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.

Yes, you read that right. The European Commission and Zuckerberg are on the same page when it comes to digital competition.

That complexity can make my brain hurt. In the ongoing lobbying around new digital competition rules (looking at you, United Kingdom), the playbook often relies on claiming such legislation places regulators too squarely at the heart of business decisions of some of the world's largest tech companies. "It's killing innovation!," comes the claim. "Officials should keep their noses out of our business!"

I have some sympathy for that argument, especially when it comes to so-called ex ante regulation, or policy efforts to curb unfair dominance before a firm becomes too entrenched in a digital market. But I can also see a massive upside for consumers if a non-Apple product/service works as effortlessly as an in-house device designed in Cupertino.

For what such competition decisions lead to, we only have to look at a previous European Commission ruling to force the iPhone maker to switch all of its devices over to USB-C technology. Apple executives equally met that 'common charger' ruling with derision. But now, USB-C is the de fault global standard, allowing one cable to connect everything from iPhones to Samsung tablets.

It's still unclear if the recent Apple decision will lead to US pushback after the White House threatened retaliatory tariffs on countries/regions that went after American tech firms. But beyond the iPhone maker, many US companies remain supportive of this specific European Commission competition decision — mostly because it's good for their own business interests.

What I'm reading

— A subcommittee of the US Senate Committee on the Judiciary will hold a hearing on the "Censorship Industrial Complex" on March 24. Watch along here. A counterpoint to that subcommittee's focus.

— Company responses to the White House's call for input on a "AI Action Plan." Palantir. OpenAI. Alphabet. Microsoft. Frontier Model Forum. Anthropic. If anyone has seen Meta's submission, please let me know.

— AI Now gives the European Commission a report card on tech for the Berlaymont Building's first 100 days. More here.

— The UK regulator Ofcom outlined what companies must now do after a deadline passed for firms to conduct illegal harms risk assessments. More here.

— Small AI language models offer a cheap option for indigenous communities to take advantage of this emerging technology, argue Brooke Tanner and Cameron Kerry for the Brookings Institution.

digitalpolitics.co/newsletter0…