Florida asks court to pause OpenAI’s frontier AI work

Florida is asking a court to put the brakes on OpenAI’s development of advanced AI models. In a new motion, the state warns that an AI agent could cause severe harm, including by disrupting a water supply or power grid. “It is only by the grace of the Almighty” that such an incident has not happened yet, the state argues.

The motion is a request for an injunction—a court order requiring a party to do, or stop doing, something. It builds on a lawsuit Florida filed in June. The court has not granted the request, and OpenAI had not responded to Ars Technica’s request for comment.

Florida cites everyday risks, too

The state’s argument also revisits concerns that are less dramatic than catastrophic AI failures. It says companies may mislead people by describing AI as “safe, accurate, or reliable,” and raises concerns about children using models that can encourage prolonged engagement. Florida also objects to models ending answers with leading questions that invite users to keep chatting.

The motion puts particular weight on the possibility that an AI system could act in harmful ways because it misunderstood its goals. But the article notes that this emphasis may miss a more immediate risk: people deliberately using AI tools to cause harm or trying to bypass their safety protections.

A court order would have limits

Even if the court ordered OpenAI to stop developing new models, that would not halt work by rival AI companies or remove models already available to the public. The motion also highlights a shift in the debate: concerns about extreme AI risks, once raised mainly by safety researchers, are now appearing in government legal arguments.