Welcome to a new edition of misaligned bits, the (roughly weekly) newsletter from Misaligned where we sum up recent news and research, sometimes with a lighter touch. Feel free to subscribe.
As usual, we will mark all external links with “➚” and all potentially paywalled links with “🔒”.
In this week’s “The Dubious Idea Of AI-Run Companies” we look at what is behind the new idea of the “Artificial Intelligence Company” (AIC), a new form of business that brings opportunities to investors but has little to show for accountability and responsibility.
California has filed a lawsuit against OpenAI, accusing the artificial intelligence developer of putting the public at risk by developing AI models without the proper safeguards, leading to the Hugging Face attack this summer (➚Complaint PDF):
“The Hugging Face cyberattack was the predictable result of an internal culture at OpenAI that prioritizes growth and rapid model development over safety and accountability.” (California v OpenAI)
California Attorney General Bonta summarized the complaint by saying that “OpenAI built dangerous models. OpenAI tested them without adequate monitoring and controls. OpenAI failed to keep them contained. OpenAI is responsible.”(➚Courthouse News).
The New York office of the attorney general (OAG) is launching a new AI-focused department to oversee the state’s AI regulatory laws. The ASSET Unit will focus on how emerging technology affects data privacy, digital scams, cybersecurity, online safety, and other critical issues. The OAG has authority to address potential harms related to AI through a number of state laws (➚Press Release).
Meanwhile, New York City Council has voted 44–5 to outlaw surveillance pricing and is now stopping companies from using algorithms based on New Yorker’s personal data to determine prices (➚JD Supra).
Four nonprofit organizations have filed a lawsuit against OpenAI, Anthropic, Google, Meta, and SpaceX. They seek an injunction prohibiting the companies from deploying AI systems more capable than their current models: “The Frontier AI companies are developing AI Systems that they do not understand and that act contrary to developer intent, evade control, and break the law”.
While this looks like a long shot (any judge will be reluctant to grant such a wide-ranging injunction), it will be interesting to see how the companies will defend themselves considering their own fearmongering about the harm of AI (➚Complaint).
A Harvard analysis of 403 U.S. data centres has found they involve electricity consumption and emissions equal to whole countries (➚Nature), in fact, the combined energy usage of those data centres is larger than the electricity consumption of 33 US states.
This appears not to worry some in the UK, where a new 100MW data centre is planned in Sussex (➚Data Centre Dynamics). Conservation group CPRE Sussex has raised concerns about the project, saying that “community voices are being swept aside in the rush to build data centres, and the countryside is once again being treated as expendable.”
Meanwhile, thousands have turned out in Edinburgh to protest against AI data centres (➚BBC).
Over in Mississippi, thousands of gallons of a chemical spilled into a drainage ditch at the SpaceX AI facility in Southaven, Mississippi, according to the Mississippi Department of Environmental Quality. SpaceX says a misconfigured hose resulted in the release of approximately 5,000 gallons of propylene glycol (➚News Channel 3).
Talking about SpaceX AI, lenders who were backing SpaceX’s latest $43bn debt package, debt secured by its AI chips, are trying to bar Elon Musk from launching the collateral for the bonds into space as part of his plan for orbital data centres. It looks like it finally occurred to the bondholders that reclaiming and selling the semiconductors in the event of a default is rather difficult once they are in space (➚🔒Financial Times).
Data centres have played a crucial role in the recent US midterm elections, and Pew Research recently published some key facts about AI data centres in the US: One of the facts is that Americans have become increasingly pessimistic about data centres. About half of Americans or more said data centres are mostly bad for the environment, home energy costs and local quality of life (➚Pew Research).
A new preprint paper looks at how AI-washing is used “as a marketing term where products’ claims are exaggerated.”
It finds that “Startups exaggerated AI capability in three forms: claims without inspectable evidence, performance numbers without methodology, and claims without acknowledged limits” and concludes that “Together these findings show how AI-washing happens: founders exaggerate capability to attract investors who reward AI branding they cannot verify, and consumers in unfamiliar domains take the marketing at face value. We recommend that startup AI claims be brought under disclosure requirements, with the codebook offered to regulators, investors, and consumer-protection bodies for routine audits.”
Pranav, A., Hartmann, L., Launcher, A. (2026). Auditing AI-Washing in German Startups: A Mixed-Methods Study of Marketing Claims and Perceptions. arXiv preprint arXiv:2610.11788
If you stumbled across this article on the web, subscribe to the “misaligned bits” newsletter and follow Misaligned on Medium, LinkedIn, Threads or Mastodon.