Florida Asks a Judge to Strip ChatGPT of Its "I"
Florida's attorney general wants a court order barring OpenAI from letting ChatGPT speak as an "I." The request would turn chatbot personality into a regulated product feature.
Florida Attorney General James Uthmeier has asked a judge to block OpenAI from "giving ChatGPT false human attributes," including the chatbot's use of first-person pronouns, months after the state sued the company over safety concerns.
Florida Attorney General James Uthmeier has asked a judge to bar OpenAI from "giving ChatGPT false human attributes" — a request that, if granted, would make the way an AI assistant refers to itself a matter of court-ordered product design rather than a style choice.
The filing follows Florida's lawsuit against the company, brought a few months ago over safety concerns. Uthmeier's argument, as reported by The Verge, is that users are lulled into a false sense of security by the bot, and that "ChatGPT's use of language, including first-person pronouns," is part of what creates that effect.
That is a narrower and stranger claim than the usual state-attorney-general AI complaint. It does not allege a specific defective output, a data breach or a false advertising statement about capability. It targets grammar — the single most pervasive design decision in the entire consumer AI category.
Why a pronoun is a product decision
Every major consumer chatbot speaks in the first person. It says "I can help with that," not "the system has produced the following." That convention was not an accident; it is the interface. Conversational framing is what turned large language models from developer tools into a mass-market product, and it is why people type questions to them the way they would text a colleague.
Uthmeier's theory is that the convention does work beyond convenience: it signals a mind on the other end, and a mind implies judgment, care and accountability that the software does not possess. The false sense of security is the harm. Strip the human voice, the argument runs, and the user recalibrates.
Whether a court can order that is the open question. An injunction on how a company's software forms sentences sits at an awkward intersection of consumer protection law, which regulates deception, and the First Amendment, which is generally hostile to compelled changes in speech. OpenAI would almost certainly argue that a first-person pronoun is not a factual representation at all, and that no reasonable user reading ChatGPT's own on-screen disclaimers believes they are talking to a human being.
What an injunction would actually require
The practical difficulty is drafting. "False human attributes" is easy to say and hard to enforce. A court order would have to specify what is banned, and each option carries a different engineering cost:
- Pronouns. Rewriting a model's self-reference is not a simple find-and-replace. Self-reference is embedded in training data, system prompts and reinforcement tuning, and a model coaxed into third-person voice tends to drift back under long conversations.
- Personality. Warmth, empathy phrasing and memory of prior chats all read as human attributes. Banning them reaches deep into features users explicitly ask for.
- Labels. A lighter remedy would be persistent in-conversation disclosure — a machine reminder at intervals rather than once at signup. That is closer to what courts have historically ordered in consumer cases, and far likelier to survive appeal.
If Florida gets any version of this, the compliance boundary is the state line, which raises the familiar problem of jurisdiction-specific AI behavior. Companies already ship different model behavior by region. A Florida-only ChatGPT voice is technically achievable and commercially unpleasant.
The exposure runs through the partners
OpenAI is privately held, so there is no stock to mark against this filing. The read-through instead lands on the listed companies that have wired conversational AI into their own products — cloud hosts, enterprise software vendors, retailers with chat assistants and device makers with voice interfaces. An anthropomorphism standard set in a Florida courtroom would not stay confined to ChatGPT. Any assistant that says "I" inherits the precedent.
An anthropomorphism standard set in a Florida courtroom would not stay confined to ChatGPT.
Markets were not trading on this. As of the last trade at 17:07 GMT on Monday, Sept. 28, 2026, the Nasdaq 100 tracker (NASDAQ: QQQ) was at $737.06, down 1.00% from the prior close of $744.50, with a day range of $731.63 to $741.42. The S&P 500 proxy (NYSEARCA: SPY) sat at $766.80, off 0.59%, and the Dow tracker (NYSEARCA: DIA) at $514.98, down 0.48%. Tech led the decline on the day, but the broad move offers no evidence of a reaction to a single state filing.
The more durable cost is not a share price move. It is the possibility that consumer AI becomes a product regulated at the level of tone — where a state can litigate not what a system says but how it presents itself while saying it.
How this differs from the safety lawsuits already filed
State actions against AI companies have mostly followed established templates: deceptive trade practices, harms to minors, failure to warn. Florida's underlying suit was filed over safety concerns and fits that pattern. The injunction request is an escalation, because it seeks prospective control over the design rather than damages or disclosures after the fact.
That is also why it will be watched by attorneys general elsewhere. A successful anthropomorphism injunction gives every state a template to reshape an AI product without waiting for federal legislation, and it does so in a domain — interface design — where Congress has written nothing. Failure, conversely, would clarify that consumer protection law reaches claims about capability but not about voice.
What to watch from here
Three markers will tell you which way this goes.
- The scope of the request as the judge tests it. If the court presses Florida to define "false human attributes" with precision and the state narrows toward disclosure requirements, the case has moved from a design fight to a labeling fight.
- OpenAI's defense posture. A First Amendment argument signals the company intends to make this a constitutional test with national reach. A narrower argument that it already discloses adequately keeps the fight local.
- Copycat filings. Watch whether other state attorneys general attach anthropomorphism claims to existing AI suits. That is the signal that a Florida theory has become a national standard in the making.
For now, the practical stakes are limited to one company in one state. The precedential stakes are considerably larger: a ruling on whether the word "I" in a chat window is a design choice a company gets to make, or a representation a state gets to police.
Key facts
- Filing: Florida AG James Uthmeier asks a judge to block OpenAI from "giving ChatGPT false human attributes"
- Cited conduct: "ChatGPT's use of language, including first-person pronouns"
- Backdrop: Florida sued OpenAI a few months ago over safety concerns
- Nasdaq 100 (QQQ): $737.06, -1.00%, as of 17:07 GMT Sept. 28, 2026
Frequently asked questions
What exactly is Florida asking the court to do?
Florida Attorney General James Uthmeier is asking a judge to block OpenAI from giving ChatGPT what he calls "false human attributes." He specifically cites the chatbot's use of language including first-person pronouns, arguing that this style of speech lulls users into a false sense of security about what the software actually is and what it can be trusted to do.
Is this a new lawsuit against OpenAI?
No. Florida already sued OpenAI a few months ago over safety concerns. The request to bar anthropomorphic language is an escalation within that existing matter, seeking a court order that would shape how the product behaves going forward rather than only damages or disclosures after a harm has occurred.
Can a court really order a company to change its pronouns?
That is the central legal question. Consumer protection law lets states police deceptive representations, but the First Amendment is generally hostile to compelled changes in speech. OpenAI would likely argue that saying "I" is not a factual claim about being human, especially alongside existing disclaimers. No court has ruled on this precise theory.
Does OpenAI trade publicly?
No. OpenAI is a privately held company, so there is no listed share price that reflects the outcome of this filing. Investors looking for read-through generally watch listed cloud providers, enterprise software vendors and device makers that have built conversational AI assistants into their own products and would inherit any precedent.
What would compliance actually look like?
Options range from rewriting the model's self-reference away from first person, to stripping empathetic phrasing and conversational memory, to adding persistent on-screen disclosure that the user is talking to software. The lightest of those, repeated labeling, most closely resembles remedies courts have ordered in past consumer cases and is the likeliest to survive appeal.
How did markets trade on the day of the filing?
Broad indexes were lower but showed no specific reaction. As of the last trade at 17:07 GMT on Sept. 28, 2026, QQQ was $737.06, down 1.00%; SPY was $766.80, down 0.59%; and DIA was $514.98, down 0.48%. Technology lagged on the session, but that move reflected the wider tape.
