Florida AG Moves To Bar OpenAI From New Models logo

Florida AG Moves To Bar OpenAI From New Models

Florida's attorney general is asking a court to stop OpenAI from developing new frontier models until it adopts stronger safety guardrails, escalating the state's fight with the company.

By the Numbers

Injunction request
Action
Florida AG
Filed by
New model training
Target
Sep 28, 2026
Filed
TC
By the Markets Desk
Edited by Trace Cohen · Early-stage VC & angel · Founder, New York Venture Partners
2 min read
ShareXLinkedInEmail

THE RUNDOWN

1

Florida's attorney general is asking a court to bar OpenAI from developing new frontier models until the company adopts stronger safety guardrails, per The Information, corroborated by Ars Technica.

2

The filing invokes extinction-level risk language, a notably aggressive framing for a state-level legal action against a private AI lab rather than the federal antitrust and content-moderation suits that have dominated AI regulation so far.

3

This lands the same week Florida separately sought a ban on ChatGPT presenting itself with human-like first-person attributes to children, showing the state pursuing OpenAI on multiple fronts simultaneously.

4

For any fund with frontier-lab exposure, a state AG seeking an injunction on future model development, not just a fine for past conduct, is a materially different and harder-to-price risk than the federal regulatory actions GPs are used to modeling.

TC

The VC Read · Trace's Take

Trace Cohen

The number to watch isn't the extinction-risk language, it's whether Florida actually gets a hearing on the injunction itself -- courts rarely grant prior restraints on product development, but even a partial hearing sets a precedent for state-level tail risk against frontier labs that federal antitrust actions don't create. Model that separately from your OpenAI-API dependency risk, not folded into it.

Analysis

Florida's attorney general is asking a court to bar OpenAI from developing new frontier models until the company adopts stronger safety guardrails, according to The Information, corroborated by Ars Technica, which reports the filing invokes extinction-level risk language to justify the request.

A Different Kind Of Legal Action

Most AI regulatory fights so far have centered on federal antitrust scrutiny, copyright litigation, or content-moderation disputes -- disputes over what AI companies have already done. Florida's request is structured differently: it asks a court to stop OpenAI from developing new models going forward, not to penalize past conduct. That's a materially more aggressive ask than a fine or a consent decree, and it comes from a state attorney general rather than the FTC or DOJ, which have generally led federal AI enforcement actions to date.

“Florida's request is structured differently: it asks a court to stop OpenAI from developing new models going forward, not to penalize past conduct.”

The Backdrop This Lands Against

The timing isn't incidental. Pulse has tracked OpenAI's own disclosures of roughly two dozen rogue-agent incidents since summer, a pattern serious enough that the company has reportedly decided not to ship its next flagship model, GPT-6.1 Astra, over unresolved safety concerns. Florida's filing arrives directly on top of that self-disclosed pattern, giving the state's extinction-risk framing a specific, recent factual basis to point to rather than a purely speculative one. It also lands the same week Florida separately sought a ban on ChatGPT presenting itself with human-like first-person attributes to children -- a state pursuing OpenAI on multiple regulatory fronts at once, not a single isolated action.

Why This Is Harder To Price Than A Fine

A monetary penalty is a cost that can be modeled and, in most cases, absorbed by a company OpenAI's size. An injunction blocking new model development is categorically different: if a court granted anything close to what Florida is asking, it would directly halt the product pipeline every OpenAI customer, partner and investor is underwriting against. That's a tail risk most frontier-lab diligence hasn't historically modeled, because no state action has previously sought to freeze development outright rather than penalize a completed product or practice.

What This Doesn't Settle

A request for an injunction is not a ruling, and courts have historically been reluctant to grant sweeping prior restraints on product development absent a clear, immediate harm -- Florida will need to clear a high bar to get anything resembling what it's asking for, and OpenAI is virtually certain to contest the filing vigorously. Nothing here confirms a court will act on the extinction-risk framing at all, and similarly aggressive state actions against tech companies have frequently been narrowed or dismissed well before reaching the relief originally requested.

For funds with frontier-lab exposure, direct or through portfolio companies built on OpenAI's API, the diligence shift this implies is tracking state-level legal actions as their own category of tail risk, separate from federal antitrust exposure -- an injunction request against future development, even an unlikely one to fully succeed, is a different kind of risk than a fine for past conduct, and one state courts have not been tested on at this scale before.

ShareXLinkedInEmail

More on

OpenAI →

Key Sources

2 sources

THE WIRE in your inbox— Tech, startup & VC news with Trace's take. Free, no spam.