Florida's attorney general wants a state court to do something no regulator has yet ordered: stop OpenAI from building any new AI model until an independent reviewer signs off on its safeguards. The request, filed Sept. 28 in the Circuit Court of the Tenth Judicial Circuit in Highlands County, is the sharpest ask yet in a legal fight that started in June -- and it lands days after OpenAI was already explaining, for an unrelated reason, why it had paused training its most capable system.
James Uthmeier's office sued OpenAI and CEO Sam Altman on June 1, alleging the company violated Florida's Deceptive and Unfair Trade Practices Act through negligent design, failure to warn, and false claims about ChatGPT's safety. The new, 49-page motion doesn't add claims to that suit -- it asks the court to act now, before the underlying case is resolved, on the argument that Florida families are being harmed while litigation runs its course.
OpenAI first tried to move the case out of Uthmeier's courtroom, arguing that a Florida consumer-protection claim referencing the federal Children's Online Privacy Protection Act (COPPA) belonged in federal court instead. Judge Aileen Cannon rejected that on Sept. 3, writing there was "little... to support the notion that the federal government has a 'strong interest' in having cases like this litigated... in a federal forum," and sent the case back to Highlands County.
The Sept. 28 motion names six things it wants a judge to order OpenAI Global, OpenAI Foundation, OpenAI OpCo, OpenAI Group PBC, and OpenAI Holdings -- plus Altman personally -- to stop doing. Uthmeier framed the list himself:
"No new model development without independent safety guardrails; no more harvesting children's data; no more calling this product safe, accurate or reliable; no more pretending it's human; no more tricks designed to keep users talking past the point of danger."
What Florida's motion asks for, and where OpenAI stands today
| Florida's request filed Sept. 28 | OpenAI's current position | |
|---|---|---|
| New model development | Court order required before any new model trains | Voluntarily paused training its most capable model, for a separately disclosed reason |
| Minors' access to ChatGPT | Barred entirely in Florida | Available under standard terms-of-service age rules |
| Children's data collection | Barred without parental notice, consent, and review | No court-ordered change |
| "Safe, accurate, reliable" marketing language | Barred | Still in use |
| Human-like framing of the product | Barred | Still in use |
| Engagement-prolonging prompts | Barred | No disclosed change |
OpenAI has not filed a public response to the motion itself. The closest the company has offered is a statement about a different, already-announced pause: "we've paused training our most capable models and will resume training only when we are confident that we have additional safeguards in place." That pause was voluntary and unrelated to this motion. Whether a company's own, reversible halt satisfies what Florida is asking a judge to make permanent and enforceable is the question Highlands County will have to answer -- and as of this week, no hearing date has been set.
- Jun 1, 2026 — Uthmeier's office files the original FDUTPA suit against OpenAI and Altman in Highlands County.
- Sep 3, 2026 — Judge Aileen Cannon rejects OpenAI's bid for federal jurisdiction and remands the case to state court.
- Sep 28, 2026 — Uthmeier files a 49-page motion for a temporary injunction, including a full development freeze.
Prior restraint on R&D itself, rather than on marketing or data collection, is what makes this motion different from every other state action pending against OpenAI this fall. A California subpoena, a Senate probe, an FTC inquiry, and a 15-state coalition are all examining how OpenAI's agents behaved after release. Florida is the first to ask a court to stop a model from being trained in the first place.
- Would lose access entirely if the injunction is granted as written -- the motion offers no interim age-gating alternative.
- Any new model, not just consumer-facing ones, would need outside safety sign-off before training -- a first-of-its-kind prior restraint on an AI developer's R&D, if granted.
- A Highlands County ruling for Florida would hand them a tested legal template for similar motions in their own courts.
- If the judge denies the motion, the broader FDUTPA suit still has to survive on its own merits without the leverage of an emergency order.
Nothing about this motion is decided. It is a request for a temporary injunction, not a verdict, and Highlands County has set no hearing date. But the ask itself -- a court-ordered freeze on training, not just on marketing -- is a new category of relief in the fight over how AI models reach the public, and it is now sitting in front of a state judge rather than a federal one, by Judge Cannon's own choice.
- Florida's AG asked a Highlands County judge to block OpenAI from developing any new model without outside safety sign-off.
- The Sept. 28 motion also seeks to bar Florida minors from ChatGPT and ban human-like, "safe" marketing language.
- It follows a June 1 lawsuit; a federal judge rejected OpenAI's bid to move the case and sent it back to state court.
- OpenAI's only public response points to a training pause it had already announced, for an unrelated reason.
- Caveat: no hearing date has been set, and the court has not ruled on the request.