Florida moves to halt OpenAI over grid, water risks

Florida's new motion builds on a broader public safety case it originally filed in June, which included concerns about false advertising around claims that AI is "safe, accurate, or reliable," dangers of letting children access addictive and sycophantic models, and worries about models soliciting engagement with leading questions. The new filing adds extinction-level and utility-sabotage-level risks to that foundation.

The motion leans heavily on so-called "misalignment" — the risk that an out-of-control AI agent might conclude that killing a cancer patient also kills cancer, or that humans won't have so many mouths to feed if they are all chemically sterilized. The state also invokes what it calls "the greatest public nuisance ever created." That framing, though, tends to overlook the practical dangers of human users intentionally misusing AI agents or actively disabling safety guardrails.

Even a court order against OpenAI would not touch models already deployed to the public, and frontier models still in training may have less well-developed guardrails than those already released. The outcome likely hinges on whether the court treats the state's extinction-level claims as actionable or focuses on the lower-level public safety arguments — and on how OpenAI responds, which the company has not yet done. The move suggests the legal and public policy mood around AI has shifted in recent weeks and months, as governments start taking seriously warnings that AI safety researchers have made for years.