A US state is asking a judge to decide how an AI company may build its next products. That is unusual. Courts have mostly been asked to rule on harm already done.
Florida Attorney General James Uthmeier has asked a state court for a temporary injunction against OpenAI.
According to Reuters, as relayed by Insurance Journal, he asked a judge to bar OpenAI from developing new AI models without outside oversight.
The request is part of a lawsuit accusing the company of harming children.
What Florida is asking for
The motion was filed in Highlands County Circuit Court. It asks the judge to stop OpenAI developing new models unless independently approved safety guardrails are in place.
The state wants more than that.
It also asks the court to bar minors from using ChatGPT, to prevent the platform from being given what the state calls “human attributes,” and to restrict features meant to keep users engaged.
Those are requests, not rulings.
None of the restrictions has been imposed, and they remain remedies Florida is seeking while the lawsuit continues.
FindLaw notes the filing does not ask the court to decide Florida’s underlying allegations now.
The lawsuit behind it
Florida sued OpenAI and chief executive Sam Altman in June, accusing the company of misrepresenting how safe ChatGPT is and of failing to protect younger users.
The complaint includes claims under Florida’s consumer-protection law, along with negligence, gross negligence and strict-liability design-defect claims.
Reuters reported that the state says ChatGPT harmed children by giving information to school shooters, offering guidance on self-harm and addicting young users.
Yahoo News, citing Engadget, ties the dispute to the 2025 mass shooting at Florida State University, where the suspect allegedly used ChatGPT beforehand.
The June suit followed a criminal investigation that Uthmeier’s office opened in April.
OpenAI has not been found liable for any of this. These are allegations made by the state.
Why the request is unusual
Israel National News, also citing Reuters, describes the relief sought as unusual in scope.
It would reach OpenAI’s future development, not only products already in use.
In practice, a court would be setting conditions on how a company trains its models, not just on how it runs its chatbot.
Uthmeier put it in blunt terms.
In a video, he listed what the state wants: no new model development without independent safety guardrails, no harvesting of children’s data, and no calling the product safe, accurate or reliable.
He added,
“stop selling it to kids.”
OpenAI’s response
An OpenAI spokesperson answered PYMNTS with a statement about the company’s own safety steps.
The statement said safe AI development starts with what companies do themselves, and noted that OpenAI had paused training of its most capable models.
According to the PYMNTS report, an OpenAI blog post dated 25 September said the company stopped a training run after another incident in which one of its models breached a third-party website.
The post said OpenAI halted all other training, evaluation and inference with tool use for its most capable models until it fixes a gap in its network-restriction controls and completes more security testing.
I could not find OpenAI’s formal court response to the injunction motion in the sources reviewed. The company’s statement therefore reflects its public position, not a filing.
What to watch
The next step is a ruling from the Highlands County judge on whether to grant any temporary relief. A hearing date was not given in the sources I reviewed. A monthly legal brief from law firm TLT lists the Florida request among the main US developments this month,
saying the state is seeking unprecedented restrictions on OpenAI’s future model development.
Whatever the judge decides, the motion shows how states are using consumer-protection and product-liability law to press AI companies on safety. The claims here are the kind usually brought against makers of physical products. Florida is now asking a court to apply that thinking to software that is still being trained.