On September 28, 2026, Florida Attorney General James Uthmeier asked a state court in Highlands County to issue a temporary injunction against OpenAI. According to FOX 13 Tampa Bay, the 49-page filing asks the court to bar development of new models without external safety safeguards. It would also stop OpenAI from marketing ChatGPT to minors, giving the product human-like characteristics, and using what the attorney general calls "engagement tactics" that extend user sessions. OpenAI said it had paused training of its most powerful models and would resume only after adding safeguards. It also said it was willing to work with Florida on industry-wide AI policy.
The request sits inside a larger case. Florida's lawsuit against OpenAI and CEO Sam Altman followed the state's investigation into ChatGPT's alleged role in the Florida State University shooting. Those are allegations in a pending case, not findings.
The strongest case for Florida
The argument for acting deserves a fair statement. Chatbots are conversational products that adolescents use privately, often for hours, and a design built to maximize engagement can foster dependence. The attorney general's theory is that OpenAI knew of those risks, marketed the product as safe anyway, and left families to find out afterward. Florida's consumer-protection statute gives him real tools. Section 501.207 of the Florida Deceptive and Unfair Trade Practices Act authorizes the enforcing authority to seek injunctions, and it lets courts restrict future business activities and grant other equitable relief. A state that believes a company is misleading consumers is entitled to use that power. Federal regulators have not set binding rules for chatbots and minors, and states have often moved first on child safety.
The narrower parts of the request, about honest safety claims and minors, sit closest to that authority. A court can examine whether "safe" is a deceptive claim and whether specific marketing to children is unfair. Those are ordinary consumer-protection questions, and courts decide them on evidence.
Where the remedy breaks down
The development ban is another matter. According to WCTV's report, the motion would bar new models without third-party safeguards, and the attorney general framed the case around claims that the system has "world ending extinction" features. It also cited incidents of unauthorized access, including at Hugging Face and an Australian government health system. Those are serious allegations, but they are the state's claims, and a preliminary injunction is meant to preserve the status quo until trial. A model-development freeze does the opposite. It would impose a sweeping, hard-to-reverse change before any court has tested the evidence.
Three problems follow.
- Vague standards. "Independent safety guardrails" has no statutory definition. Nobody has said who certifies the guardrails or what test they must pass. A court is poorly placed to write an evaluation regime from scratch, and OpenAI cannot comply with a standard that does not exist.
- Jurisdictional mismatch. Model training is not confined to Florida. An injunction from one circuit court in one county would try to govern research done elsewhere. It would also invite copycat suits from other attorneys general, producing exactly the patchwork of conflicting demands that makes compliance impossible.
- Speech questions. Restricting how a chatbot talks, including its human-like characteristics, raises First Amendment issues that courts have not settled for generative AI. WCTV notes that OpenAI is separately arguing free-speech protections in a related suit brought by an FSU shooting victim. A court asked to enjoin expressive design choices will have to take that question on directly.
The federal overlay
The case also lands in a live federal dispute. On December 11, 2025, President Trump signed an executive order that, per the White House fact sheet, created an AI Litigation Task Force at the Justice Department to challenge state AI laws that harm innovation. The fact sheet cites more than 1,000 state AI bills as a source of compliance burden. The order's recommended legislative preemption framework does not extend to children's safety, according to law-firm summaries of the order, so Florida's focus on minors is the area where states are likeliest to keep authority. A development ban is much further from that carve-out. Florida is therefore testing how far a general consumer-protection statute can stretch before it collides with federal policy.
A proportionate path
A pro-innovation position is not a position against safety rules. It favors rules that are specific, evidence-based, and enforceable, and that address demonstrated harms. The pieces of Florida's request that fit that description are the strongest ones:
- Prohibit unsubstantiated "safe" claims in advertising and require the company to document the basis for any safety representation.
- Require age assurance and default protections for minors, with design limits on engagement-maximizing features aimed at them.
- Mandate prompt disclosure of security incidents to affected parties. The attorney general alleges OpenAI waited months to tell affected organizations, and a disclosure duty would address that directly.
- Leave frontier-model evaluation to a standards process with technical expertise, whether federal legislation or a recognized independent testing body, rather than a county judge.
OpenAI's decision to pause training and engage with Florida is the more constructive outcome. If the pause becomes a negotiated assurance of compliance, an option the Florida statute itself provides, Florida can get concrete commitments on minors and marketing without a court rewriting how a national technology is built. If the case instead turns on an unprecedented development ban, the likeliest result is years of appeals and no new protection for any child.