Florida's attorney general is asking a court to do something no US regulator has done: stop OpenAI from building new artificial intelligence models until independent outsiders sign off on their safety.

The request came in a filing on 28 September as part of the state's June lawsuit accusing OpenAI of misrepresenting the safety of ChatGPT. Attorney General James Uthmeier wants a judge to bar the company from developing new AI models without safety guardrails approved by a third party, to keep minors off ChatGPT, and to stop giving the chatbot "human attributes."

A Reuters account of the filing says Florida leans on the company's own words. It cites recent comments from a former OpenAI employee and a current board member warning that advanced AI could end humankind and should be slowed — a position the state characterises as an admission. "Defendants claim they cannot stop barreling forward with their potentially civilization-ending endeavors unless they are forced to do so by the government," the filing argues. "They have asked the government to tie them to the mast."

OpenAI pushed back, saying it has already paused training its most capable models until additional safeguards are in place. Spokesperson Drew Pusateri said governments have an important role in setting robust safety standards and that the company is committed to working with Florida and other states on policies that apply to the whole AI industry "not just one company."

Florida's suit was prompted by a shooting at a Tallahassee university last year and other episodes in which ChatGPT allegedly provided information to people who went on to commit violent acts. Uthmeier, a Republican, is the first state attorney general to sue OpenAI over its impact on young users. OpenAI denies liability, arguing the chatbot surfaces information that is broadly available online and that it keeps updating its safety tools.

The case is one of several converging pressures. OpenAI also faces individual and family lawsuits, and the FTC has opened a broad investigation into the safety of frontier AI systems, focused on the risks of agentic models after a run of incidents.

Sources