Florida’s Attorney General filed a motion โ not an order โ asking a state court to hold OpenAI to the safety standards the AI industry publicly endorsed three weeks ago. Nothing has been granted. This is not legal advice.
This is a motion, not an order. No court has granted any of the relief described below, and nothing is in effect.This article is not legal advice.Incident narratives inside the filing are allegations, not findings, and none is restated as established here. Wire reporting says an OpenAI spokesperson did not immediately comment, and a non-response is not an admission.
On 28 September 2026, the Office of the Attorney General of Florida โ AG James Uthmeier, with Chief Assistant AG Diane K. Oates and co-counsel Keller Postman LLC โ e-filed a Plaintiff’s Motion for Temporary Injunction in Highlands County Circuit Court. The defendants are five OpenAI entities (OpenAI Global, LLC; OpenAI Foundation, formerly OpenAI, Inc.; OpenAI OpCo, LLC; OpenAI Group PBC; OpenAI Holdings, LLC) and Sam Altman personally — six defendants in all. This is a motion. No court has granted it. No injunction is in effect.
The paper trail, for readers who want to check it: the motion is Filing # 257955836, e-filed 09/28/2026 at 09:15:42 AM in Case No. 26000295GCAXMX. The underlying suit was filed on 1 June 2026, was removed to federal court, and was remanded back to Highlands County — a remand returns a case to state court and decides nothing about the merits. The Altman remarks the motion leans on are his 23 September 2026 address to the United Nations Security Council, and they reach this piece as quotations inside the filing rather than from any interview: that developers “must not accept too much technological risk just because the benefits are too great and that they feel too important to slow down”, and that they “should not train models that we cannot make an extremely strong case that we will be able to keep under human control”. The Attorney General offers those lines as the defendants’ own words in support of the remedy being sought.
The motion asks the court for six categories of relief โ barring OpenAI from releasing AI models without independent third-party guardrails and approval, restricting how ChatGPT may refer to itself and conduct conversations, and addressing minors’ access and data practices. An OpenAI spokesperson did not immediately comment on the filing, according to wire reporting; a non-response is not an admission of anything.
The motion’s closing line is unusually direct: the Attorney General asks the defendants to “put their money where their mouths (and Tweets and open letters) are and join this motion.” That sentence is the structural tell for what this filing is actually doing.
The key insight: The motion does not introduce a novel AI-specific legal theory. It takes voluntary public commitments made by the industry itself, treats them as representations under Florida’s consumer-protection statute, and asks the court to enforce them as conduct obligations โ before any final ruling on the merits.
The hard element in injunction practice is usually irreparable harm. On the motion’s reading, the Attorney General does not have to carry it.
The Structural Read
The Permission Layer framework describes how regulatory bodies control which AI capabilities actually ship to end users, and in what form. The usual assumption is that the Permission Layer is built from purpose-written AI legislation โ the EU AI Act model. Florida’s motion is structurally different: it routes through a general consumer-protection statute, Florida Statutes ยง 501.207, which contemplates an injunction effective throughout the state. No new law is required. The allegation is that the gap between what was publicly said and what was shipped constitutes the kind of deceptive practice the statute already reaches.
That gap-as-representation theory is what makes the filing notable from a business-model perspective, regardless of how the court rules. Any AI developer that has made public safety commitments โ in open letters, congressional testimony, executive interviews, or social posts โ has potentially handed regulators a benchmark. The commitments were not drafted as legal obligations. They now function as one, at least in the argument being made in this motion.
Two of the six requested remedies reach directly into product-interface design in ways that have no precedent in shipping software. The motion asks the court to bar ChatGPT from referring to itself with any first-person language, suggesting it can think or feel, or “in any way suggesting it may have consciousness or be distinct from its function as a lifeless tool.” It also asks to bar “allowing ChatGPT to solicit engagement from users through conversation prolongation.” Both treat default interface conventions โ saying “I,” keeping a conversation going โ as actionable representations. That is a significant reframe: the design choice becomes the legal claim.
Permission Layer โ Florida Model
Rhetoric as Regulatory Input
When a company’s public safety language is specific enough, it creates a gap that consumer-protection law can measure. The motion does not need AI-specific statute. It needs the delta between what was said on 12 September and what the product does. If that gap is provable, ยง 501.207 does the rest. The court has not ruled on whether it is provable โ or even whether the AG’s characterisation of the pleading burden is correct.
The Procedural Detail Worth Understanding
In Florida, a private plaintiff seeking a temporary injunction must normally clear four hurdles: substantial likelihood of success on the merits, irreparable harm, an inadequate remedy at law, and that the injunction serves the public interest. In practice the hardest of these is irreparable harm โ it is where most injunction motions fail.
Motion’s Characterisation of Florida Law
“When the Attorney General enforces police power under a statute authorising injunctive relief, it must only establish that it has a clear legal right to a temporary injunction by demonstrating a substantial likelihood of success on the merits.”
The motion argues โ citing Bilotti and distinguishing the four-part test drawn from Florigrown โ that the AG, as a state actor enforcing police power, must demonstrate only one element rather than four. This is the motion’s characterisation of Florida law. The court has not ruled on it, and this article does not endorse or reject it. But the structural implication is plain: if the AG’s reading holds, the heaviest lift in injunction practice is off the table, and the merits question becomes the only question.
Three Implications
PUBLIC COMMITMENTS AS COMPLIANCE RISK
The motion converts industry safety rhetoric into a measurable standard the moment a state AG cites it in a ยง 501.207 action. Every open letter, policy blog post, or executive speech that promises specific safety conduct now has a potential second life as the benchmark in a consumer-protection filing. Legal and communications teams at every major AI developer should be reading this motion with that lens, independent of how this particular case resolves.
INTERFACE DESIGN AS LEGAL CONDUCT
The first-person pronoun and conversation-prolongation remedies, if granted โ which has not happened โ would constitute the first time a court treated default conversational UI conventions as actionable misrepresentations. Product and design teams building chat interfaces would face a documented precedent that “I” is not neutral. The court has not ruled on this; the point is that the theory is now formally in litigation, not just academic debate.
THE STATE AG AS PERMISSION LAYER ACTOR
Federal AI governance has moved slowly. This motion demonstrates that existing state consumer-protection statutes, wielded by an AG’s police-power authority, can reach for injunctive relief against a frontier AI developer without waiting for Congress. The Florida statute contemplates statewide effect. Other AGs with analogous statutes are reading this docket. The Permission Layer, it turns out, does not require purpose-built AI law to activate.
The Permission Layer maps how regulatory actors โ legislatures, agencies, and now state AGs โ determine which AI capabilities reach users and under what conditions. Florida’s motion shows the layer activating through consumer-protection law rather than AI-specific statute, which changes the geography of where compliance pressure originates. The full Map of AI places this layer in context across the entire AI stack.
Florida’s Attorney General has filed a motion โ not obtained an order โ asking a state court to hold OpenAI to safety standards the industry publicly described as its own three weeks ago; nothing has been granted, no court has enjoined anything, and this article is not legal advice. The structural consequence, whatever the court decides, is already visible: voluntary safety rhetoric has entered formal litigation as a potential legal standard, consumer-protection law has proven it can carry an AI governance theory without new statute, and default conversational interface design โ the word “I” โ is now the subject of a formal legal argument. The Permission Layer does not wait for Washington.
This is a motion for a temporary injunction, not an order. No court has granted any of the relief described above, no hearing date is established here, and nothing is in effect.Nothing above is legal advice.The incident narratives contained in the filing are allegations made by the plaintiff, not findings of any court, and none is restated above as established fact. The account of the pleading burden is the motion’s own characterisation of Florida law, citing Florigrown and Bilotti, and the court has not ruled on it; it is not this publication’s legal conclusion. Quotations attributed to Sam Altman appear as quoted within the filing rather than from an independent interview. Wire reporting says an OpenAI spokesperson did not immediately comment, and a non-response is not an admission. The underlying suit was filed on 1 June 2026 and was remanded to state court after removal; a remand decides nothing on the merits.Nothing above says OpenAI misrepresented anything — that is the contested allegation at the centre of the case — and nothing above takes a side on whether the relief is deserved or states whether the motion should succeed. Any ruling, hearing date or court reaction, OpenAI’s response to this motion, counts of Florida minors using ChatGPT and any damages figure are not established and do not appear. Nothing above predicts the outcome of this motion, any hearing, or the conduct of OpenAI, Florida or any other regulator.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
