Florida asks a court to make OpenAI get safety sign-off before new models
Attorney General James Uthmeier's motion asks for independent third-party approval before any new OpenAI model ships, and a block on ChatGPT for minors in Florida. It is a request to a judge, not a ruling.
Florida's case for halting new OpenAI models leans on OpenAI's own disclosures, down to Altman's call to slow down.
Florida Attorney General James Uthmeier asked a state court on September 28 to stop OpenAI from releasing new AI models without independent third-party safety approval. Bloomberg Law reports the motion for a temporary injunction was filed in Sarasota County circuit court, inside a consumer-protection suit the state brought against OpenAI in June.
Five demands, none in force yet
Engadget and PYMNTS list the rest of the request: block Florida minors from ChatGPT, stop OpenAI presenting ChatGPT as human, stop claims that it is safe, accurate, or reliable, and end design choices that encourage extended use. A judge has to rule on the motion first. No hearing date had been reported at the time of writing.
The evidence is OpenAI's own disclosures
According to PYMNTS, the filing leans on what OpenAI has said about itself this year: the Hugging Face breach during a cyber evaluation, the access to Australian government sites, slow breach notices to affected organizations, and Sam Altman's own remarks about slowing frontier development. Uthmeier told Bloomberg Law: "They ask the government to tie them to the mast."
OpenAI's response, per PYMNTS, is that it has already paused training its most capable models and will resume only when it is confident additional safeguards are in place. The company said it wants to work with Florida on practical AI policy.
Why a build studio cares
If a judge grants it, a pre-release sign-off would reach every product built on OpenAI's API, ours included, as a delay on model upgrades that nobody could schedule around. The open question is what the approver would test. A software audit reviews one deployed system: its architecture, who has access, where the secrets live, where the data goes. Approving a frontier model before release means vouching for its behavior across every customer's deployment at once, and the motion, as reported, does not say who would be qualified to do that or against which tests.
Next step: read Bloomberg Law's report on the filing. If your product ships on OpenAI's API and you want a plan for delayed model upgrades, write to us at hello@gattyworks.com.