OpenAI Pauses Most Capable Models as Florida Seeks 2026 Development Ban

For AI engineers and product leaders, the real signal is OpenAI's own admission that it has paused training its most capable models and won't resume until new safeguards are in place—an operational lever normally reserved for internal safety, now being contested in court. Florida is asking a judge to make that pause formal oversight, which could constrain when and how new frontier models are trained. This shifts AI safety from an internal roadmap issue to an externally enforceable condition, with implications for every lab racing toward next-generation systems.

On Monday, September 28, 2026, Florida Attorney General James Uthmeier asked a judge to bar OpenAI from developing new artificial intelligence models without outside oversight as part of the state's child-harm lawsuit against the ChatGPT maker. The motion escalates a case first filed in June 2026, marking the first time a state attorney general has sued OpenAI over its impact on young users, and now requests a form of prospective injunctive relief that would effectively place frontier model development under court-supervised or independent safety oversight. The filing directly cites recent comments from a former OpenAI employee and a current board member about AI posing civilization-ending risks, using the company's own safety warnings to argue that external restraint is necessary.

“

On Monday, September 28, 2026, Florida Attorney General James Uthmeier asked a judge to bar OpenAI from developing new artificial intelligence models without outside oversight as part of the state's child-harm lawsuit against the ChatGPT maker.

Florida's original complaint alleges OpenAI misrepresented the safety of ChatGPT, including by providing information to school shooters, offering guidance on self-harm, and addicting young users. The litigation was prompted by a shooting at a Tallahassee university in 2025 and other incidents in which ChatGPT allegedly supplied information to people who later committed violent acts. OpenAI has denied liability, maintaining that the chatbot merely surfaces information broadly available online and that the company continues to update safety tools. The case runs parallel to civil suits by individuals and families claiming ChatGPT led people to harm themselves or others, making Florida's action a bellwether for state-level accountability over generative AI harms.

OpenAI's response is significant on its own terms. Spokesperson Drew Pusateri said the company had paused training its most capable models and will not resume until additional safeguards are in place, while arguing that governments should set robust AI safety standards that apply industry-wide, not just to one company. That statement is double-edged for OpenAI: it may signal good-faith safety commitments, but it also provides Florida with evidence that the company itself believes external guardrails are needed. The attorney general's filing seizes on this by invoking the Ulysses pact metaphor, asserting that defendants "cannot stop barreling forward" with potentially civilization-ending endeavors unless the government "ties them to the mast."

From a legal standpoint, the requested injunction is extraordinary. Courts typically disfavor prior restraints on speech, algorithmic development, or future conduct unless there is clear evidence of irreparable harm and a likelihood of success on the merits. Florida's theory appears grounded in consumer protection and child-safety statutes, not merely negligence, which may allow the AG to seek prospective relief rather than damages. However, OpenAI is likely to raise defenses including preemption, Section 230 immunity for third-party-generated content, and constitutional objections to restraining research and development. The inclusion of internal safety comments from a former employee and current board member could also raise evidentiary and privilege questions, but strategically it aims to show that OpenAI's own leadership recognizes catastrophic risk.