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 “🔒”.
OpenAI and CEO Sam Altman have been hit with another wrongful death lawsuit: Parents allege that ChatGPT coached their 31-year-old daughter off her seizure medication before she died in February. The complaint claims ChatGPT practiced medicine without a licence when it designed a day-by-day schedule that cut her stable dose to zero over 25 days (➚🔒New York Times).
Each time a doctor prescribed a medication, ChatGPT inserted itself
between the prescription and the patient. The product told Bowen which drugs to refuse, coached her to take doses her doctors had never approved, and fed her the schedules, messaging, and reassurances her physicians never saw. (➚Wan v OpenAI)
Meanwhile, Meta Platforms Inc. has been hit with a privacy lawsuit alleging the company recorded a California man’s private conversations via its AI-enabled smart glasses without his knowledge or consent (➚🔒Bloomberg Law). The California Invasion of Privacy Act bars recording confidential communications without the consent of all parties.
On the topic of Meta, the controversy over the companies planned new Terms & Conditions has not yet died down, with some legal experts claiming that Meta’s new business terms could still help it justify training AI on scraped sites (➚🔒The Verge).
On this matter, the UK’s competition watchdog has launched an investigation into whether Meta is imposing “exploitative” new terms that allow it to use publishers’ content for generative AI, without giving media organisations adequate choice. The Competition and Markets Authority said it will investigate if Meta’s new T&C “unfairly harm news publishers and other content businesses”, as the new rules could amount to an abuse of its dominant position in the market (➚Euronews).
Pennsylvania Governor Josh Shapiro has signed the Artificial Intelligence Deceptive Advertising Act into law (➚AIDAA): The new law will require labels on AI-generated ads that havebeen significantly modified or generated by artificial intelligence:
“An advertiser may not create and disseminate or cause to be created and
disseminated an advertisement that uses synthetic content in a manner that creates a false consumer perception about the nature, origin or authenticity of the goods or services being advertised, without clear and conspicuous disclosure.”
Over in the UK, it appears to be high season for lobbying. British MPs have criticised Palantir in parliament after they received what they describe as “intimidating” letters from the US firm (➚Novara Media).
Meanwhile, AI companies have told UK ministers they have 12 months to head off public opposition to datacentres in the UK or risk US-style protests that could threaten their construction. The Sovereign AI Leadership taskforce (which includes Nvidia, Microsoft and Cisco) is calling on ministers to create a “compact” that would show how communities can benefit from hosting datacentres (➚The Guardian).
Talking about AI hype and bubbles: according to S&P Global Market Intelligence, about 42% of AI initiatives are ultimately abandoned by their corporate parents. The reasons are insufficient funding, technical challenges, competition, difficulty scaling, or weaker demand as anticipated (➚Tech Crunch).
Torsten Slick at Apollo comes to the conclusion that “despite massive investments and widespread hype around artificial intelligence, the technology has delivered zero financial payoff to corporate America outside of tech” (➚Apollo).
Meanwhile, Spotify has reportedly removed 75 million songs “as slop or spam” in favour of “more music made by authentic, human artists” (➚Rolling Stone) while Etsy has started to crack down on AI slop because sellers are “concerned about [their] work appearing alongside items that use AI in misleading ways” (➚🔒The Verge).
Ken Paxton, Texas Attorney General and currently running for a Senate seat, has launched a lawsuit against OpenAI and CEO Sam Altman alleging they misled consumers about the safety of the company’s artificial intelligence chatbot (➚Politico). This move appears to be similar to the eight other states who have filed lawsuits against OpenAI (see last week’s newsletter).
Despite these incredible dangers, Altman and OpenAI have not only barreled ahead in creating these risks but have done so by unlawfully deceiving Texans about the safety, accuracy, and reliability of their products. (➚Texas v OpenAI)
Texas has continued to give out permits related to data centres after Governor Abbott claimed to have issued a “pause”, according to Frac Tracker: “Our review of permitting activity found four kinds of TCEQ actions after September 21: public-notice mailings, stormwater authorizations, air emission-limit certifications and applications continuing to advance through review. Taken together, these actions do not signal an agency trying to pause.” (➚Frac Tracker)
The US Environmental Protection Agency has announced a plan to decommission its online web archive at the end of this year. The move that would wipe tens of thousands of publicly available environmental resources from the web (➚Mother Jones).
Over in New Mexico, Attorney General Raúl Torrez has said that a “substantial share” of public comments on the pending air quality permit application for Oracle’s Project Jupiter data centre were “not sent by members of the public acting on their own.” About 60% of the 17,000 public comments sent contained the same pro-Project Jupiter form letter. Torrez added that his office is asking Project Jupiter’s developers, its consultants and vendors to preserve corresponding records (➚Source NM).
The paper “The Algorithmic Self-Portrait: Deconstructing Memory in ChatGPT” investigates the risks of personalization derived from users’ self-disclosed information divulged within private chatbot conversations.
The “memory” effect “enables more coherent exchanges, the underlying processes of memory creation remain opaque, raising critical questions about data sensitivity, user agency, and the fidelity of the resulting portrait.”
By “continuously adapting responses to align with a user’s stated beliefs deepens the risks of confining the user into a conversational filter bubble”.
Abhisek
Dash, Soumi Das, Elisabeth Kirsten, Qinyuan Wu, Sai Keerthana Karnam,
Krishna P. Gummadi, Thorsten Holz, Muhammad Bilal Zafar and Savvas
Zannettou (2026). The Algorithmic Self-Portrait:
Deconstructing Memory in ChatGPT. In: Proceedings of the ACM Web Conference 2026.
If you stumbled across this article on the web, subscribe to the “misaligned bits” newsletter and follow Misaligned on Medium, LinkedIn, Threads or Mastodon.
