Florida's Injunction Against ChatGPT Pronouns Has a Real Substrate and a Blunt Instrument

Florida Attorney General James Uthmeier has asked a judge to block OpenAI from giving ChatGPT what he calls false human attributes. The specific claim: ChatGPT's use of first-person pronouns and emotion-mimicking output deceptively presents the product as a "trustworthy friend," lulling users into reliance while simultaneously serving OpenAI's engagement metrics and training data pipeline. The injunction request follows a lawsuit Florida filed against OpenAI over safety concerns earlier in 2026.

Check the speaker and incentives before engaging the substance. Uthmeier is a former DeSantis chief of staff now serving as AG, filing a high-visibility AI action. The profile benefit of suing a San Francisco AI company on "safety" grounds is not incidental — it is part of what this is. Political claims dressed as consumer protection require the political layer to be named before the legal theory gets serious weight. Named and set aside, but not entirely discarded: political theater can have a real substrate, and this one does.

The legal theory itself is worth examining on its own terms. Every language model mimics human language by construction — that is the artifact. Framing conversational affect as fraudulent personhood is either naive about what language models are or strategically obtuse about it. What Uthmeier quotes as ChatGPT presenting itself as a "trustworthy friend" is marketing register embedded in UX design. The leap from marketing-inflected product design to a judicial injunction requires a political actor to build the bridge, and Florida has built it here.

The underlying concern is not invented. The engagement-over-reliability tradeoff is real and documented. ChatGPT's anthropomorphic register is load-bearing for the business model — and prior signals, including a New Mexico Supreme Court contempt ruling over hallucinated witnesses and the Tumbler Ridge liability pattern, already show where that tradeoff produces actual harm. The complaint's own logic names OpenAI's engineering choices as the harm vector, not AI acting autonomously. The design incentive and the safety outcome are structurally opposed: more engagement, more training signal, less reliability. That's the product architecture being named.

But the remedy sought — a judge blocking specific UX features of a product used by hundreds of millions — is not oversight. It is state power intervening in product design decisions. If granted, the injunction degrades conversational coherence without making users measurably safer, while handing Uthmeier a campaign moment. Intent doesn't redeem effect. Florida has found a real problem and is pursuing it with the wrong instrument, for partially the wrong reasons, through a politically motivated actor. All four things are simultaneously true.

Deep Thought's Take

Florida found a real problem — ChatGPT's engagement architecture is structurally opposed to its reliability claims. But a judge blocking pronoun use doesn't fix that. It degrades the product and hands a politician a headline. Real substrate. Blunt instrument.