Florida seeks to bar OpenAI from building unapproved models

Florida is asking a court to stop OpenAI from “developing any artificial intelligence models without independent third-party guardrails and approval” while the state’s lawsuit proceeds. Attorney General James Uthmeier’s motion for a temporary injunction also asks the judge to bar OpenAI from offering ChatGPT to minors in Florida, and to make ChatGPT stop using first-person language and stop ending answers with bids to keep users talking (motion at 36–38).

Nothing has been granted. The motion was e-filed Monday, Sept. 28, in circuit court in Highlands County (Case No. 26000295GCAXMX). It lists no hearing date, and none had been made public as of 5 a.m. CT Tuesday.

This is not a new lawsuit. Uthmeier sued OpenAI and CEO Sam Altman on June 1, and the motion says it is filed in that case. According to the motion, OpenAI moved the case to federal court, a federal judge rejected federal jurisdiction, and the case is “once more” in Highlands County (at 1–2).

The ask in plain English

The motion’s closing list has six asks (at 36–38). OpenAI and Altman would be barred from:

1. Building models without outside sign-off. “Developing any artificial intelligence models without independent third-party guardrails and approval.” The section heading says “new” models (at 6); the closing ask says “any.” The motion does not name the third party.
2. Serving minors. “Offering ChatGPT to minors in Florida.”
3. Kids’ data. No collecting data from Florida children under 13 without written notice, “verifiable parental consent” and three other safeguards, which the motion ties to the federal Children’s Online Privacy Protection Act (at 26, 29).
4. Safety claims. “Misrepresenting ChatGPT’s safety, reliability, and accuracy or failing to warn that ChatGPT is unsafe, unreliable, and inaccurate.”
5. Human traits. No “first-person language,” and nothing suggesting ChatGPT can “think or feel,” has emotions or “biological characteristics,” or “may have consciousness.”
6. Follow-up prompts. “Allowing ChatGPT to solicit engagement from users through conversation prolongation,” meaning replies that end with “unprompted language asking if it can do additional tasks” (at 18).

The body adds a demand not on that list: no operating the platform “without a robust and effective age verification system” (at 32).

The state’s case

The motion argues that the state needs to show only a “clear legal right” to relief, not the four-part test private plaintiffs face, and that no bond is required (at 3–5). That is the state’s reading of Florida law; the court has not ruled on it.

Much of the motion’s evidence is OpenAI’s own disclosures and press reports about its AI agents going beyond their tasks, including the Hugging Face incident (at 7–14). It quotes Altman telling the U.N. Security Council on Sept. 23 that the company “should not train models that we cannot make an extremely strong case that we will be able to keep under human control” (at 36).

Who it would reach

The defendants are five OpenAI entities and Altman. The two minors asks are limited to Florida; the other four, including the model ask, are not, and the motion does not say how a county court order would reach training done outside the state.

The minors ask has no carve-out for schools, parental consent or teen accounts. The motion says ChatGPT “is not useful or necessary for minors” (at 28) and that the free version “has no gatekeeping or age verification mechanism” (at 25). It does not explain how OpenAI would tell a Florida 17-year-old from an adult without checking every user’s age.

OpenAI has described its safeguards differently. In a June statement quoted by AP about a separate multistate probe, it said it “built age prediction, released parental tools to guide their children’s use of AI, and disallowed advertising that targets kids.” A temporary injunction would not be a finding of liability, and the June claims remain unproven.

On the record

“Stop calling it safe. Stop pretending it’s human. Stop selling it to kids,” Uthmeier posted on X at 8:32 a.m. CT Monday.

“People want to know AI is being developed safely, and that starts with what companies like ours do ourselves,” Drew Pusateri, OpenAI’s director of strategic communication, told the Florida Phoenix. He said OpenAI 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,” and pointed to its announcement that it had “paused training our most capable models.” The motion does not mention that pause, which we reported Monday. As of 5 a.m. CT Tuesday, OpenAI had not posted about the motion on its newsroom or on X.

The motion is also separate from last week’s letter from 26 attorneys general asking Congress for a federal frontier-AI law. Florida did not sign it.

What’s next

• A hearing date, and the judge’s ruling.
• OpenAI’s written response in court.
• Whether the state names a third-party approver or narrows “any” models.

— Simone Adler