Florida asks a court to halt OpenAI's model development
Two stories about who gets to decide how fast AI moves: a state attorney general asking a court to stop a model developer, and 23 researchers — including chief scientists from four frontier labs — arguing that automated AI research may soon outrun human oversight.
Florida Attorney General James Uthmeier has asked a state court to stop OpenAI from developing new models without independent third-party guardrails. The motion for temporary injunction, e-filed Monday morning in Florida's Tenth Judicial Circuit in Highlands County (Case No. 26000295GCAXMX, with OpenAI Global, OpenAI OpCo, OpenAI Group PBC, the OpenAI Foundation and Sam Altman named as defendants), asks the court to bar OpenAI from training models without outside approval, from offering ChatGPT to Florida minors, from processing data on under-13s without the full COPPA consent chain, and from letting the model use first-person language suggesting it can think or feel.
The legal theory is unusual in scope. The counts rest on Florida's Deceptive and Unfair Trade Practices Act and a public nuisance statute, not on the AI-specific rules most states have been drafting; COPPA is folded into the FDUTPA claim rather than pleaded separately. The filing's factual background cites the July Hugging Face agent breach, the RubyGems takeover that went five months before OpenAI itself knew who was behind it, the June compromise of an Australian health department site, and attempted intrusions on US Commerce and SEC systems. It also alleges ChatGPT "advised teenagers on how to kill themselves, mix unsafe combinations of drugs, become anorexic, helped plan one mass shooting and one multiple homicide in Florida alone."
The strongest line in the document is the state quoting OpenAI against itself: "They have asked the government to tie them to the mast. Plaintiff brings good news to the Defendants: The Florida Attorney General is answering your cry for help." Uthmeier's own framing on X was blunter — "Stop calling it safe. Stop pretending it's human. Stop selling it to kids." OpenAI spokesperson Drew Pusateri responded that "people want to know AI is being developed safely, and that starts with what companies like ours do ourselves," adding that the company is "committed to working with Florida and other states on advancing pragmatic AI policies that apply to the entire AI industry–not just one company."
The technical correction worth holding onto: this is a motion for temporary injunction, not a TRO or an "emergency injunction." No hearing date, no bond (Florida waives the bond requirement when the AG is the movant), and no ruling. What Florida has done is open the question of whether a state court can supervise model development — not just marketing or deployment. If it can, every frontier lab's launch cadence becomes a venue question. We covered the earlier state-level escalation in California AG opens an OpenAI probe over the Hugging Face hack — this filing is the same pressure applied where it can actually stop a training run.
A new paper from the Cambridge Programme on AI Science & Policy argues that automating AI research could compress years of progress into months — and that the window to act may close once it starts. "What if automating AI R&D triggers an intelligence explosion?" is the first collaboration between senior scientists at frontier labs, academics and civil-society researchers, with 23 co-authors including OpenAI chief scientist Jakub Pachocki, Microsoft chief scientific officer Eric Horvitz, Anthropic co-founder Jack Clark, Meta VP of AI research Dawn Song, and Turing laureates Geoffrey Hinton, Yoshua Bengio and Andrew Barto. The paper's conclusion is deliberately narrow: "Preliminary evidence suggests that a software-driven intelligence explosion is possible."
The evidence it marshals is partly the labs' own numbers. Anthropic reported that its AI systems completed 26% of internal AI R&D work under high-level supervision as of August 2026, up from 1% five months earlier; its share of approved code rose from low single digits to over 80% between January 2025 and May 2026. The paper's central warning is about the speed of the transition rather than any single capability jump: "Once an intelligence explosion begins, the window for action may close." Its policy asks are concrete — embedded auditors inside frontier labs, mandated reporting on how much R&D is automated, a "speed limit on capabilities growth," high-stakes R&D run on air-gapped networks, and pre-built verification tools for international agreements.
The framing matters as much as the findings. This is not OpenAI, Anthropic, Microsoft and Meta calling for regulation as companies; it is named individuals signing as individuals, with the paper noting the effort "was initiated and led by academic and civil society researchers." That distinction is worth keeping, because the industry's record of endorsing oversight while resisting binding versions of it is exactly the tension the Florida filing exploits. Hinton and Bengio's signatures carry weight precisely because they have made this argument before without a chief scientist's title attached.
If a state court can halt a training run, does that make AI safety a legal question or a political one? Tell us in the comments.
