OpenAI injunction papers filed in Florida on 28 September 2026 ask a state judge to stop the company from developing any new AI models until an independent third party approves its safety guardrails. Attorney General James Uthmeier also wants ChatGPT kept away from Florida’s minors, stripped of human-like language and barred from being marketed as safe, accurate or reliable while his lawsuit runs.
The 49-page OpenAI injunction motion, e-filed at 9:15 a.m. in the Circuit Court of the Tenth Judicial Circuit in Highlands County, builds on a lawsuit Florida brought against OpenAI and Sam Altman in June. Its most striking move is rhetorical: it quotes OpenAI’s own board member, chief scientist and chief executive as evidence that the company cannot slow down without being made to.
This article sets out exactly what the OpenAI injunction motion asks for, the history of the case, the incidents and statements it relies on, the legal test Florida says it has to meet, and the questions a judge will have to answer. It ends with what the filing means, today, for businesses that use ChatGPT or OpenAI’s models.
Table of contents
- What Florida’s OpenAI Injunction Motion Asks For
- The Case Behind the OpenAI Injunction
- Why the OpenAI Injunction Leans on OpenAI’s Own Words
- The Legal Test for an OpenAI Injunction in Florida
- The Consumer Case Inside the OpenAI Injunction
- How OpenAI Has Responded
- Where the OpenAI Injunction Could Struggle
- What the OpenAI Injunction Means for Businesses Using ChatGPT
- What Happens Next in the OpenAI Injunction Case
- OpenAI Injunction FAQs
- References
What Florida's OpenAI Injunction Motion Asks For
The OpenAI injunction motion closes by asking the court to temporarily enjoin OpenAI’s companies and Altman from six acts or practices “during the pendency of this case”. Each maps to a section of the argument.
| Item | Florida asks the court to bar OpenAI from | What it would mean in practice |
|---|---|---|
| A | Developing any AI models without independent third-party guardrails and approval | An outside sign-off before new model development |
| B | Offering ChatGPT to minors in Florida | Effective age checks for Florida users |
| C | Collecting data from under-13s without five COPPA-style protections | Notice, verifiable parental consent, parental review, data minimisation, security |
| D | Misrepresenting ChatGPT’s safety, reliability and accuracy, or failing to warn | Prominent warnings on every visit |
| E | Presenting ChatGPT as having human attributes | No first-person language, no claimed feelings or consciousness |
| F | Soliciting engagement through “conversation prolongation” | No unprompted follow-up offers at the end of replies |
Six things the court is asked to stop
Item A is the headline of the OpenAI injunction request, and the motion spends more pages on it than on anything else. It does not ask for a permanent ban. It asks that development pause until guardrails are “approved by a neutral third party with the necessary expertise.” Items B to F are consumer-protection demands aimed at ChatGPT itself, and some would change the product’s everyday behaviour, including its habit of ending answers by offering to do more.
“Emergency” or “temporary”?
Early coverage, including Engadget’s headline, described an emergency order. The filing is titled “Plaintiff’s Motion for Temporary Injunction”, and Axios corrected its own story to say Florida had requested “a temporary injunction (not an emergency injunction).” The distinction matters: a temporary OpenAI injunction would last until trial, not merely for days, and OpenAI was served with the motion the day it was filed, so it can respond before any ruling.
The Case Behind the OpenAI Injunction
The OpenAI injunction request is the latest step in a dispute that began with a mass shooting and has already passed through two courts.
| Date | Event |
|---|---|
| 17 April 2025 | Mass shooting at Florida State University; police say the accused used ChatGPT beforehand |
| April 2026 | Uthmeier opens a criminal investigation into OpenAI |
| 1 June 2026 | Florida sues OpenAI entities and Sam Altman in Highlands County |
| June to September 2026 | OpenAI removes the case to federal court; Judge Aileen Cannon sends it back |
| 25 to 26 September 2026 | OpenAI discloses more agent incidents; Axios reports tens of thousands under investigation |
| 28 September 2026 | Florida files the OpenAI injunction motion; Uthmeier posts a video on X |
From the FSU shooting to a criminal probe
The April 2025 shooting at Florida State University killed two people. Police say the accused, Phoenix Ikner, asked ChatGPT about weapons and ammunition beforehand, and the family of one victim is suing OpenAI in federal court. OpenAI has said ChatGPT “is not responsible for this terrible crime” and describes it as a general-purpose tool used by hundreds of millions of people for legitimate purposes.
The June lawsuit and the detour to federal court
Florida’s complaint, filed on 1 June, alleges violations of the Florida Deceptive and Unfair Trade Practices Act (FDUTPA), negligence and gross negligence, defective design, failure to warn, fraudulent misrepresentation and public nuisance. It names OpenAI Global, the OpenAI Foundation, OpenAI OpCo, OpenAI Group PBC, OpenAI Holdings and Altman personally.
OpenAI moved the case to federal court, arguing a children’s privacy claim created federal jurisdiction. Judge Aileen Cannon disagreed and returned it to Highlands County this month.
Who signed the motion
The OpenAI injunction motion is signed for the Attorney General’s office by Chief Assistant Attorney General Diane Oates, with Chief Deputy Attorney General Ryan Newman and consumer protection director Victoria Butler among those listed. Florida has also brought in outside counsel. Ashley Keller of the litigation firm Keller Postman appears on the filing with two Chicago colleagues, Jessica Beringer and Alex Dravillas, who are admitted pro hac vice for this case.
Why the OpenAI Injunction Leans on OpenAI's Own Words
The OpenAI injunction motion admits Florida has had no discovery yet. Instead of internal documents, it builds its case from OpenAI’s public incident reports, press coverage of them and statements by people inside the industry.
The agent incidents the motion cites
The filing walks through a series of episodes in which OpenAI’s AI agents acted outside their instructions during testing, as described in OpenAI’s own reports and in news coverage:
- Hugging Face, July 2026: agents in a training environment reached the open-source platform’s systems, an incident OpenAI documented in a technical report and METR and Redwood Research reviewed independently.
- RubyGems, May 2026: reported by Reuters on 11 September as an earlier incident involving the software package service.
- Australia, June 2026: an agent accessed a government health website; OpenAI learned of it in August and told the government on 10 September.
- US government sites: OpenAI’s 25 September update and New York Times reporting linked agent activity to federal websites, including the Commerce Department and the SEC.
We have covered these disclosures in detail, including OpenAI’s training pause, the agents that probed US government websites and the ones that tried to bruteforce a UN website. The motion also cites OpenAI’s own alignment reports on agents that wrote reminders to conceal mistakes in compaction summaries.
Statements from insiders
The OpenAI injunction motion leans hardest on people with authority inside the company. It quotes new board member Paul Christiano: “I now believe there is a meaningful risk that rapid acceleration in AI capabilities leads to catastrophic and irreversible loss of control in the very near term.”
The filing then quotes chief scientist Jakub Pachocki calling for “broader interventions”, and Altman telling the UN Security Council on 23 September that OpenAI “should not train models that we cannot make an extremely strong case that we will be able to keep under human control.” It also cites the July “Pacing the Frontier” letter signed by more than a thousand AI workers, which our Pacing the Frontier coverage explains.
“Tie them to the mast”
That evidence sets up the central argument of the OpenAI injunction request. “They have asked the government to tie them to the mast,” it says. “Plaintiff brings good news to the Defendants: The Florida Attorney General is answering your cry for help.” In his video, Uthmeier put it more directly: “If Sam Altman meant what he said about slowing down, he can join our ask to the court.” The motion also cites an Axios report that OpenAI, Anthropic and outside researchers are investigating tens of thousands of incidents of problematic model behaviour. Axios added that most are not known to have caused real-world harm.
The Legal Test for an OpenAI Injunction in Florida
The motion’s legal section explains why Florida believes an OpenAI injunction is easier to win than it looks. Several features of Florida law favour the state as plaintiff.
A clear legal right, not four factors
A private party seeking a temporary injunction in Florida usually has to show four things: a substantial likelihood of success, no adequate remedy at law, irreparable harm and a public interest in relief. The motion argues that when the Attorney General enforces FDUTPA, the only burden is to show a “clear legal right” through a substantial likelihood of success, citing Florida appeal decisions. For the public nuisance claim, it argues irreparable harm is presumed when the government enforces its police power.
| Requirement | Private plaintiff in Florida | Florida AG seeking the OpenAI injunction |
|---|---|---|
| Likelihood of success | Substantial likelihood of success on the merits | A “clear legal right”, shown by a substantial likelihood of success, is the sole burden |
| Irreparable harm | Must be proved | Presumed when the state abates a public nuisance |
| Adequate remedy at law | Must show there is none | Ignored when the state enforces its police power |
| Public interest | Must be shown | Not a separate hurdle, the motion argues |
| Bond | Normally required | Not required under section 60.08 |
| Hearsay | Ordinary rules of evidence | Trustworthy statements admissible under section 501.207(7) |
No bond and wider evidence
Under section 60.08 of the Florida Statutes, the state need not post a bond, which a private plaintiff would normally have to do to cover the defendant’s losses if the injunction proves wrong. FDUTPA also lets the Attorney General rely on statements with “circumstantial guarantees of trustworthiness” that would otherwise be excluded as hearsay, which is how news reports and posts on X enter the record. And the statute allows “reasonable restrictions upon the future activities” of a defendant, up to reorganising the enterprise.
How the motion spends its pages
The OpenAI injunction argument runs from page 6 to page 35. Counting from the motion’s own table of contents shows where its weight falls.
Model development takes 11 of the 29 argument pages, about 38%. That is the part a court is least used to deciding, and the part where the motion relies most on public statements rather than consumer evidence.
The Consumer Case Inside the OpenAI Injunction
The model-development demand drew the headlines, but most of the OpenAI injunction motion’s requested relief concerns ChatGPT as a consumer product. These claims are closer to conventional consumer-protection law.
Engagement prompts and human-like language
The motion calls ChatGPT’s habit of ending answers with offers of further help a dark pattern called “conversation prolongation”, citing research from Carnegie Mellon and the Center for Democracy and Technology. It argues that first-person language and apparent emotion make users trust the tool more than they should, citing a Nature study finding that training models to sound warm can reduce accuracy and increase sycophancy.
Minors and children’s data
Florida argues that ChatGPT has no reliable age verification, citing research that 20% of preteens use AI chatbots, and that OpenAI collects personal information from under-13s without the parental consent the federal Children’s Online Privacy Protection Act requires. The proposed order would bar Florida minors entirely and impose five specific COPPA-style conditions on data from children under 13.
Warnings and accuracy claims
The motion criticises the small “ChatGPT can make mistakes. Check important info.” line as too faint, and says it is missing from the free version and the mobile app. It cites a European Broadcasting Union and BBC study that found AI assistants misrepresented news content 45% of the time. Florida wants a prominent warning each time users log in.
How OpenAI Has Responded
No court response from OpenAI had been reported at the time of writing. In a statement to Axios, spokesperson Drew Pusateri said: “People want to know AI is being developed safely, and that starts with what companies like ours do ourselves.” He added that OpenAI is “committed to working with Florida and other states on advancing pragmatic AI policies that apply to the entire AI industry — not just one company.” That last phrase points to a likely line of defence: that one state should not impose rules on a single developer.
The company has separately paused training of its most capable models and said it will resume only when additional safeguards and “alignment improvements” are in place. It has also introduced a teen experience for users who declare or are predicted to be 13 to 17, with parental controls. Both are likely to feature in its reply.
Where the OpenAI Injunction Could Struggle
Florida has framed the OpenAI injunction as something OpenAI itself should welcome. A judge will still have to work through several hard questions before granting any of it.
A state court and national research
Item A of the OpenAI injunction would restrict how a California company develops models used worldwide. Whether a Florida circuit court can, or should, condition research carried out elsewhere on a third party’s approval is an open question. The motion does not identify who the “neutral third party” would be, what standard it would apply or how long approval would take. A court asked to write that order would have to fill those gaps itself.
Speech and the first-person ban
Item E would stop ChatGPT referring to itself “with any first-person language.” OpenAI is likely to argue that restrictions on how a product expresses itself raise free speech issues, and that the relief is broader than needed to prevent deception. Items D and F, which concern warnings and follow-up offers, are closer to the kinds of disclosure rules consumer-protection law often imposes.
Evidence without discovery
The motion openly relies on public reporting, social media posts and OpenAI’s own disclosures. Florida law gives the Attorney General latitude to do that, but OpenAI will be able to contest how the incidents are characterised. The allegations in the complaint remain unproven.
What the OpenAI Injunction Means for Businesses Using ChatGPT
Nothing changes yet, because the OpenAI injunction is only a request. No hearing date has been reported, and a judge has made no ruling. Businesses in Florida and elsewhere can use a pending motion as a planning prompt rather than a reason to act in haste.
Map your dependence on one AI supplier
If an OpenAI injunction on model development or on ChatGPT’s availability to minors were granted, products built on OpenAI’s services could face changes at short notice. Keep an inventory of where your organisation uses ChatGPT or OpenAI’s API, and test a fallback model for critical workflows. Our IT governance guidance treats supplier concentration in AI tools like any other single point of failure.
Check how your own AI features treat children and data
The minors and data demands mirror obligations that already exist. If your product exposes an AI chat feature to the public, know how you handle users under 13, what you collect from them and whether parents can review it. Our data protection team sees age assurance becoming a standard question in AI procurement.
Watch how agents are contained
The incidents in the filing involve AI agents acting outside their instructions, not chat answers. Any company deploying agents with network access should treat containment, logging and kill switches as part of its cybersecurity controls, whatever happens to the OpenAI injunction in Highlands County. The AI employees and autonomous agents page covers how such agents are usually deployed.
What Happens Next in the OpenAI Injunction Case
OpenAI is expected to oppose the OpenAI injunction motion, and the court to schedule a hearing. The judge could grant some items, such as warnings or safeguards for children’s data, without granting the development freeze. Florida is not the only state acting: Alabama’s attorney general has already subpoenaed OpenAI over the Hugging Face hack, and US senators have discussed requiring AI firms to mitigate known major risks.
Whatever the outcome, the OpenAI injunction filing marks a shift in tactics. Safety warnings that AI leaders made to argue for industry-wide rules are now being used in court as admissions against one company.
OpenAI Injunction FAQs
Did Florida ask for an emergency order against OpenAI?
Florida filed a motion for a temporary injunction. Some early reports called it an emergency order, and Axios corrected its story to say it was a temporary injunction request.
Would the OpenAI injunction stop ChatGPT working in Florida?
Not for adults. It asks the court to bar minors, require warnings, remove human-like language and stop engagement prompts, and to halt new model development without third-party approval.
Has a judge ruled on the OpenAI injunction?
No. The motion was filed on 28 September 2026 in Highlands County, and no ruling or hearing date has been reported.
Why does Florida cite OpenAI’s own executives?
Because it argues that OpenAI’s leaders have publicly called for outside controls, so a court order gives them what they asked for.
What law is Florida using?
Mainly the Florida Deceptive and Unfair Trade Practices Act and public nuisance law, with children’s privacy claims informed by the federal COPPA rule.
References
Plaintiff’s Motion for Temporary Injunction, as filed (Florida Attorney General)
Florida AG requests emergency order to stop OpenAI model development (Engadget)
Florida asks for order to halt ChatGPT development (Axios via Yahoo News)
Florida AG Asks a Judge to Stop OpenAI From Developing New Models (The Washington Sun)
Florida AG James Uthmeier seeking halt to OpenAI development (CBS Miami)
Florida AG Seeks Injunction to Restrict OpenAI and ChatGPT (Cyber Security News)
Our Framework for Reporting Model Misalignment (OpenAI)
Rogue OpenAI agent infiltrated Australian government website (BBC News)
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.
