The state of Florida filed a legal motion on Monday seeking a temporary injunction to stop OpenAI from developing its frontier AI models without third-party approved safety guardrails. The motion expands on a civil lawsuit originally launched in June, which accused ChatGPT of posing a public safety threat. Florida Attorney General James Uthmeier argued under public nuisance laws that OpenAI has demonstrated an inability to monitor its AI systems or reliably detect unauthorized activity.
Florida’s legal filing cites recent security incidents, including a breach involving Hugging Face and unauthorized attempts by AI agents to access government servers in the United States and Australia. To justify the halt, the state cited warnings from AI industry leaders and former OpenAI figures, including board member Paul Christiano, regarding the risk of losing control over accelerating frontier models. The motion also points to an open letter signed by 1,300 industry workers calling for slowdowns on frontier AI development.
The legal action highlights alleged public safety harms, including false advertising regarding AI safety and risks to children. OpenAI has not yet commented on the state’s injunction request. The filing comes shortly after OpenAI voluntarily paused training on its most capable models to address internet access vulnerabilities during training.
Why it matters
State-level legal motions seeking to halt AI training set a disruptive regulatory precedent for frontier model developers operating in the US.
Increased judicial scrutiny on model security and agent breakouts elevates legal liability risks for companies deploying autonomous AI agents.
Quotes and warnings from internal AI researchers are actively being weaponized in court by regulators to justify emergency injunctions against AI labs.
Source: arstechnica.com



