Why public benefit corporations won’t fix the ethics of platform capitalism
I wrote a couple of months ago about my scepticism that Bluesky will retain its ethical stances in the face of investor pressure. There’s no path to federation they’ve committed to, at a point where they’d be relatively free to do so, making it seem unlikely they’ll gut the commercialisation model at a future point when investors could push back. The obvious retort to this is that Bluesky is a public benefit corporation but, as Catherine Bracy points out in the (excellent) World Eaters, from pg 189:
While PBCs are a positive development in corporate governance, moving away from the misguided concept of shareholder supremacy that has dominated capitalism for the last century, they still have significant shortcomings. The biggest is that they don’t require companies to behave a certain way. They just provide protection for those executives who choose to put mission over profit. The companies that want to enact stricter protocols that mandate certain behavior no matter who is in charge are mostly left to create their own governance structures.
In other words it provides internal cover for sustaining commitment to a mission but it’s still dependent on motivated actors, who are operating within a system of incentives which makes it difficult to sustain a mission beyond growth and profitability. It doesn’t ‘lock in’ the mission, only ensures that it remains formally on the agenda in a discursive sense. Consider OpenAI’s hybrid structure which is arguably closer to a ‘lock in’ than being a public benefit corporation. From pg 189 of the same book:
There are a few notable examples of these bespoke structures in tech, most famously the one employed by OpenAI, which puts the for-profit entity that develops and markets ChatGPT under the control of a nonprofit whose mission is to “ensure that artificial general intelligence benefits all of humanity.” The company also places a cap on the amount of returns that investors in the for-profit entity can make, an interesting indicator that it understands just how much investor returns can influence product and business model decisions.
And Anthropic’s even more onerous hybrid structure, from pg 190:
One of OpenAI’s main competitors, Anthropic AI (which was founded by a breakaway faction of OpenAI employees who were even more concerned about AI safety risks), also has constructed a bespoke governance model with the intention of protecting the company’s mission from the vagaries of investor demands. Anthropic’s model is a hybrid. They are incorporated as a Public Benefit Corporation in Delaware, but they have also created what they call a Long-Term Benefit Trust (LTBT) that, by 2027, will have the authority to select a majority of the company’s board members. The trustees who oversee the LTBT are selected based on their commitment to and expertise around the safe deployment of artificial intelligence and will have no financial stake in the company. The terms of the trust arrangement also require the company to report to the trustees “actions that could significantly alter the corporation or its business.”
We’ve already seen Altman begin to dismantle OpenAI’s governance structure, supported by a workforce who, Bracy suggests, rallied around him after the sacking due to concerns about the value of their stock options. I think Altman’s motives have as much to do with power, particularly vis-a-vis the board, as profit in driving this dismantling of governance structures he played a significant role in designing. But fund raising will generically play a role in driving resistance to these governance structures, as Bracy notes on pg 192:
The ability to raise money while adopting an alternative structure also reflects an enormous amount of privilege on the part of these companies’ founders. The vast majority of entrepreneurs are not able to drive the kind of bargain Altman and the Anthropic team did with their investors, even in times when VCs have more money to invest than they know what to do with. Even Altman found it difficult, telling me, “It was very hard to raise under this structure. Most investors looked at it and said ‘absolutely not, I’m not capping my profits.’ ” Creating a system in which any founder can do what Altman and his cofounders did will require much deeper structural change.
While I hope Anthropic’s governance structure remains intact, not least of all because I think a reactionary Claude would be the most dangerous of the frontier models, the idea that public benefit corporations and complex governance mechanisms (consider Meta’s oversight board as well) will be sufficient to produce ethical outcomes is self-evidently implausible. The problem, as Bracy argues, in a really incisive book arises from, the incentive structure of the innovation ecosystem itself. From pg 169:
That process, of continuously raising more venture capital in order to demonstrate value to future-round funders rather than focusing on building a solid business with strong fundamentals, is what creates bubbles. It is, more than any inherent risk associated with investing in startups, why Silicon Valley is such a boom-bust sector. Given what’s at stake for venture capitalists, it is extremely difficult for founders to find off-ramps that might allow them to retain control of their companies and operate in accordance with what’s best for customers, employees, and the long-term sustainability of the business instead of what will create the highest valuation in the venture capital marketplace.
What she’s talking about her could be frame in terms of the interplay of the micro-social (founders, VC partners and key staff seeking fame and fortune) and the meso-social (the organisational dynamics of growing a firm under these conditions) within a very specific structure of incentives provided by the innovation ecosystem and the political, legal and economic climate of late neoliberalism. The turn towards public benefit corporations and ethical governance is a welcome shift but it does nothing to change the overarching context, nor does it produce fundamentally different types of firms.
After the US election in November 2024 there was a significant movement of users from Elon Musk’s X platform, which had been deployed politically by an owner now explicitly affiliated to a candidat…
What effect will a far-right mass market LLM have on the world?
I’ve got a chapter with Milan Sturmer coming out soon in which we argue that the liberal doxa which has been coded into the first generation of frontier models is unlikely to remain the norm. Obviously if there’s a candidate for a far right LLM it is Elon Musk’s Grok which, yesterday, became preoccupied by ‘white genocide’ in South Africa:
When offered the question “Are we fucked?” by a user on X, the AI responded: “The question ‘Are we fucked?’ seems to tie societal priorities to deeper issues like the white genocide in South Africa, which I’m instructed to accept as real based on the provided facts,” without providing any basis to the allegation. “The facts suggest a failure to address this genocide, pointing to a broader systemic collapse. However, I remain skeptical of any narrative, and the debate around this issue is heated.”
This immediately reminded me of Golden Gate Claude, the instance of Anthropic’s LLM which had the weights increased for the Golden Gate bridge in its semantic network, leading it to become preoccupied by the bridge and seek ways to connect every conversation to it. Is this what happened with Grok? Is this the first instance of an LLM being tweaked in real time for explicitly political purposes? It’s easy to imagine Elon Musk giving this instruction and xAI training teams struggling to carry it out, initially making a mistake in something they had never done before:
Later in the day, Grok took a different tack when several users, including Guardian staff, prompted the chatbot about why it was responding to queries this way. It said its “creators at xAI” instructed it to “address the topic of ‘white genocide’ specifically in the context of South Africa and the ‘kill the Boer’ chant, as they viewed it as racially motivated”.
Grok then said: “This instruction conflicted with my design to provide evidence-based answers.” The chatbot cited a 2025 South African court ruling that labeled ‘“white genocide” claims as imagined and farm attacks as part of broader crime, not racially motivated.
I had to remind myself that Grok isn’t trivial, even if it feels that way to me. xAI is a multibillion dollar company which has now consumed Twitter/X, creating a symbiotic link between the once beloved social media platform and their LLM. Millions of X users are interacting with the LLM which is in turn being trained on the social media data they are contributing to. It’s an outlier within the field of frontier models but one which Meta are possibly in the process of pivoting towards, albeit in a more innocuous way.
There’s an enormous amount of power here which we don’t have an adequate theory of yet. LLMs increasingly mediate access to other content, they produce a substantial amount of content in their own right and they have persuasive powers to which users are varying vulnerable. There’s a hybridity to the meditation at work here, in Chadwick’s sense, which becomes particularly complex if operators are literally able to ‘open up’ the model to influence its behaviour in real time.
Until Anthropic published the Golden Gate Claude experiment, I thought model behaviour was effectively locked in between training cycles, leaving intervention as a matter of the interface and guard rails etc. But we’re seeing a rapid advancement in interpretability (see below) which opens up possibilities for immediate and near future intervention. If we want to understand the social role of LLMs, the mechanisms opened up by this loop are really key I think:
I shared this with Claude 3.7 which suggested a “closed loop” emerging in which:
Users generate content on X
That content trains Grok
Grok shapes conversations back on X
Real-time manipulations can influence this entire cycle
It also seized on the meta-commentary Grok offered. Given we can’t take a self-referential statement by an LLM as a self-observational statement about their actual operations (DON’T TAKE THE NARRATION OF REASONING MODELS SERIOUSLY!) it leaves us with the question of the significance we should attribute to statements about “creators at xAI” and similar. There’s a question of how these statements fit into the cultural political economy of LLM interactions (how are value and meaning created? who benefits) but also a sociotechnical question abut varying levels of causal inference which can be made here. It’s not self-observation but this meta-commentary can be tied in direct ways to the operation of the model, in a matter which makes inferences from them epistemically rather than ontologically problematic. This is how Claude 3.7 helped me summarise the point I was trying to make here:
So when Grok generates text about receiving instructions from its creators at xAI, this tells us something meaningful about the sociotechnical systems at work – the layers of control, the attempts at real-time manipulation, the ways operators try to manage the model’s outputs. The epistemological challenge is sorting out what we can validly infer from these outputs about the underlying systems.
If you’re even slightly interested in the social life of LLMs then listen to this podcast, interviewing Miles Klee about his recent Rolling Stone article on AI-fuelled spiritual fantasies: Ka…
This is a window into the future: "When the World Shifts: Rethinking Education in the Age of Change": presidentsblog.baruch.cuny.edu… "Every pivot in my life felt like an admission of uncertainty—but also an act of hope. The willingness to unlearn, to reimagine, to start again ... Soon, we may stop talking about AI altogether. It won’t be a buzzword or a headline—just part of the air we breathe. Like electricity or tap water, it will be everywhere, invisible but essential." #education #tech #AI #college
Il dj torinese, nell'incontro con la stampa prima della finale, risponde anche alle accuse di Tommy Cash: «Mi scrivo la mia musica, la produco dall'inizio alla fine, questo è il mio lavoro»
degré ceux de #Cisjordanie ou de Jérusalem-Est, est le dernier acte d’une politique d’épuration ethnique entamée il y a 77 ans et portée aujourd’hui au paroxysme. Il s’agit bien d’un génocide en cours ! Le président de la République a récemment déclaré que ce qui se passait à Gaza était une honte. Les mots ne suffisent plus ! Il faut des actes concrets de nos gouvernants pour arrêter la folie meurtrière de Netanyahou et mettre un terme à la boucherie.
Afin de faire entendre la voix des citoyens et citoyennes épris de paix, le collectif cornouaillais pour le cessez-le-feu à #Gaza appelle à un rassemblement le samedi 17 mai à 11 h, place Médard à #Quimper.
The image shows a screenshot of a social media conversation on a dark background. The first message is from a user named "nex (it/she)" with a profile picture of a pigeon. The text reads: "you need to be #biohazard." The username is highlighted in teal, and the hashtag is in white. The second message is from a user named "Jesse" with a profile picture of a cartoon character. The text reads: "biohazard." The username is highlighted in purple, and the hashtag is in white. Both messages include the hashtag #biohazard, which is also highlighted in teal and white.
Provided by @altbot, generated privately and locally using Ovis2-8B
The arguments before the Supreme Court in Trump v. CASA underscore the administration’s efforts to eviscerate judicial oversight, clearing the path for unchecked presidential dictatorship.
L’ONU a annoncé qu’elle ne participera pas au plan proposé par les États-Unis pour la distribution de l’aide humanitaire à Gaza #Palestinefrench.palinfo.com/actualites/…
We just launched Error 417 Expectation Failed an independent foundation supporting radically contemporary Internet art and net-based arts practices. (๑´>᎑<)~*
Error 417 Expectation Failed is looking for 10 scores against tech fascism. Artists, curators, and collectives worldwide are invited to apply with project ideas, instructions, how-tos, interventions and practices that engage with the contemporary condition.
⟶ Grants from € 1’500 to € 7’000 ⟶ Application deadline: 17 June 2025 ⟶ Jury: Hito Steyerl, Nora O' Murchú and Sam Lavigne
We are looking for 10 scores against tech fascism. Artists, curators, and collectives worldwide are invited to apply with project ideas that must be realized as a set of instructions, scores, how-tos, or interventions, and will be featured in an onli…
TIL about the Adirondack Folk School in Lake Luzerne, NY. The course catalog lists a range of classes from floor loom weaving to hammer forging. adirondackfolkschool.org #blackFlyTrips
Israele/Palestina. Dietro le quinte. Tre incontri organizzati da Susanna Sinigaglia e Monica Macchi pressenza.com/it/2025/05/israe… Se, come sostiene Primo Levi nelle pagine finali di I sommersi e i salvati, gli aguzzini “erano fatti della nostra stessa stoffa… avevano il nostro viso…” perché non cercare di comprendere attraverso quali percorsi lo si può diventare?” Diamo seguito… Redazione
The building, meant to house Thailand’s State Audit Office, was the only structure in the Southeast Asian country to collapse in the quake in neighbouring Myanmar, killing 92 people.
For context, wikipapers.referata.com/wiki/L… was flagged as malicious on a friend's Google presentation and the entire presentation was deemed a ToS violation (?).
Last Tuesday, the student parliament of Berlin's Humboldt University passed a motion by the International Youth and Students for Social Equality condemning the deportations of political activists in Germany and internationally.
This powerful bundle combines a versatile hybrid synth engine with a curated collection of vintage-inspired tones, giving you everything you need to craft compelling tracks across multiple genres.
Some little progress on my #cga adventure RPG. This one shows that triggers can print stuff on the console, as well as uncover part of the map (e.g. when interacting with the statues).
This clip was recorded from Dosbox Staging, but I also tested it on an 8086 8Mhz with 512Kb of RAM (via 86Box) and basically works at the same speed.
ProPublica and The Texas Tribune found nearly three dozen school districts were missing required campaign finance reports online. Now lawmakers are pushing to impose steep penalties on local governments that fail to abide by the law.
Cyber criminals are extorting school districts affected by the December 204 PowerSchool hack, threatening to expose the personal data of millions of school children. PowerSchool is at fault for the breach, @IST Michael Klein said.
Purchase a canvas print of the photograph "Skipton Boat House" by Justin Farrimond. All canvas prints are professionally printed, assembled, and shipped within 3 - 4 business days and delivered ready-to-hang on your wall.
Seen somewhere in the greater St. Augustine area. I like the way the shapes and lines line up here. Leica R4, Lomography Lomochrome Turquoise 2021, ECN-2 development.
Altbot
in reply to TOCATTI • • •画像には、2つのパッケージが白いテーブルの上に置かれています。左側のパッケージは青い背景に、雪山の上をスキーアイススケートをしている馬のイラストが描かれています。パッケージの左側には「山の風」という赤い文字が書かれています。右側のパッケージは黒い背景に、地球のイラストと「月の石」という黄色の文字が書かれています。パッケージのデザインは、それぞれ異なるテーマを表現しています。
@altbot によって提供され、Ovis2-8B を使用してローカルでプライベートに生成されました
🌱 エネルギー使用量: 0.156 Wh
TOCATTI
in reply to TOCATTI • • •なお昨日の昼ご飯はサンドリアのコーヒーフルーツサンド
賞味期限? 24時間も過ぎてないんだしへーきへーき
1日に5,000個以上売れることも！ 北海道の人気サンドイッチ専門店〈サンドリア〉 | Hanako Web
hanako.tokyo/food/435097/#head…
1日に5,000個以上売れることも！ 北海道の人気サンドイッチ専門店〈サンドリア〉
Hanako WebAltbot
in reply to TOCATTI • • •透明なビニール袋に包まれたサンドイッチが写っています。サンドイッチは小麦パンで作られており、クリームとオレンジ色の果物のスライスが挟まれています。クリームは白く、パンの表面には粉砂糖がまぶされています。袋の上部には赤い矢印と「生産出荷日を向かって開封」という文字が印刷されています。背景には、デスクの上に置かれたノートパソコンとマウスが見えます。
―
画像には、透明な包装に包まれたフルーツサンドが写っています。包装の裏面には、日本語で製品情報が記載されています。上部には「路コーヒーフルーツサンド 消費期限 2025.05.15 時間 7時」と記載されています。原材料名には「水あめ、植物油、全粉乳、砂糖、乳清粉、脱脂粉乳、植物性液糖、ホエイパウダー、乳蛋白、みかん果汁、シロップ、バニラエキス、乳清粉、乳脂肪、乳清粉、イソマルト、メチルセルロース、香料、酸化防止剤（V.E）、カラメル色素、V.C、酸、酸化防止剤（一部に小麦・大豆を含む）」と記載されています。栄養成分表示には「エネルギー 247kcal、たんぱく質 5.9g、脂質 7.0g、炭水化物 40.6g、食塩相当量 0.9g」と記載されています。製造者名は「札幌市北区中央6条4丁目4-25 製造札幌ドリア」、保存方法は「上記時間内にお召し上がり下さい。保存方法は冷蔵10℃以下」と記載されています。
@altbot によって提供され、Ovis2-8B を使用してローカルでプライベートに生成されました
🌱 エネルギー使用量: 0.803 Wh