Florida asks court to stop ChatGPT from acting human and targeting kids

Florida Attorney General James Uthmeier is asking a court to ban OpenAI from giving ChatGPT human-like traits and from marketing the chatbot to minors.

The request does not stop there. Uthmeier also wants to block OpenAI from developing new models without independent safety reviews. That second demand, if granted, would inject a court-supervised review process directly into OpenAI's research pipeline — an unusual degree of judicial oversight for a private AI lab.

OpenAI, for its part, said it has paused training of its most capable models.

The Florida request targets two behaviors that have drawn increasing scrutiny from regulators and lawmakers: anthropomorphized AI interfaces, and products aimed at children. A chatbot that presents itself as human-like raises distinct concerns for younger users, who may struggle to distinguish a machine from a person. Marketing such systems to minors compounds that risk, according to the state's reasoning as reflected in the requested injunction.

The demand for independent safety reviews before new model development speaks to a different set of stakes. OpenAI's most capable models are the ones policymakers worry about most — the systems that could, in worst-case scenarios, be misused or behave unpredictably at scale. Uthmeier's filing essentially argues that OpenAI should not be trusted to police that risk on its own, and that a court should enforce external checks instead.

OpenAI's statement that it has paused training of its most capable models is notable in this context. It suggests the company has, at least temporarily, halted work on the very category of systems the Florida request seeks to constrain. Whether the pause is connected to the litigation, or to other pressures, is not specified. But the timing means the court could be weighing an injunction over a development pipeline that is currently standing still.

The case matters beyond Florida. State attorneys general have become some of the most aggressive AI regulators in the United States, filling a vacuum left by the absence of comprehensive federal AI legislation. A court order banning human-like traits in ChatGPT, blocking marketing to minors, and requiring independent safety reviews would set a precedent that other states — and other AI companies — would have to reckon with.

It would also mark one of the first times a court, rather than a legislature or an agency, imposed substantive development restrictions on a frontier AI lab. The industry has argued that self-governance and voluntary commitments are sufficient. A ruling in Florida's favor would test that position directly.

For OpenAI, the stakes are commercial as well as legal. Human-like voice and personality features have become a core part of ChatGPT's appeal, and younger users represent a significant growth market. An injunction covering both would force product changes, not just legal compliance.

What happens next depends on the court. If it grants any part of Uthmeier's request, OpenAI could face binding constraints on how ChatGPT behaves, who it is sold to, and what the company may build next — while its most advanced training runs remain on hold.