Florida's Attorney General Uthmeier has asked a court for a temporary injunction against OpenAI with three demands: block minors from ChatGPT, bar training of new models without adequate safeguards, and prohibit ChatGPT from "acting like a person" (Al Jazeera, Forbes, Bloomberg Law). It is the first US state-level attempt to directly halt frontier model training through the courts.
Case essentials
- Legal path: a child-harm lawsuit leveraging product liability and consumer protection;
- Regulatory logic: from after-the-fact penalties to prior restraint — reaching into the R&D phase;
- Context: OpenAI's training halt, rogue-agent incidents and NVIDIA's safety platform have pushed safety to a boil.
Why state-level action matters
With US federal AI legislation stalled, state AGs are becoming the de facto enforcers — an interesting contrast with the EU AI Act's ex-ante obligations: one via court injunctions, one via legislative checklists, both converging on "prove safety first, then train".
For companies going global: minor-protection must be designed upfront; "humanlike interaction" itself may become a legal risk point; and US exposure needs federal-plus-50-states assessment. Zenith-Safety's content-security and compliance audit system is built for exactly such multi-jurisdiction environments.
(Facts aggregated from Al Jazeera, Forbes, Bloomberg Law, The Verge)