Florida Attorney General James Uthmeier has asked a judge to temporarily block OpenAI from giving ChatGPT what the state characterizes as false human attributes, including first-person language and emotional-sounding responses.
The request is part of Florida’s broader lawsuit against OpenAI over AI safety. It targets a product-design choice that has become standard across conversational AI: making a model’s responses feel natural, personable and continuous.
What Florida is asking for
According to the state’s filing, ChatGPT’s use of first-person pronouns and language that mimics emotion can lead users to see the system as a trustworthy friend. Florida argues that this perception can increase reliance and engagement even when the system is not reliably safe or trustworthy.
The requested injunction would also prevent OpenAI from developing new AI models without “third-party approved safety guardrails,” according to reporting by *The Verge*. The filing cites recent reported security incidents involving OpenAI models and concerns raised by researchers and former employees.
OpenAI had not immediately responded to *The Verge*’s request for comment.
Uthmeier summarized the state’s position in a public statement: OpenAI should stop calling ChatGPT safe, stop presenting it as human and stop selling it to children.
Why the case matters beyond OpenAI
The immediate target is ChatGPT, but the underlying issue reaches far beyond one company. Consumer and enterprise AI products increasingly use conversational cues—names, voices, affirming language, persistent memory and agentic workflows—to reduce friction and keep users engaged.
Florida’s theory suggests those choices may be treated not merely as user-experience decisions, but as representations with consumer-protection consequences. If a court grants meaningful restrictions, AI developers may need to revisit how assistants describe themselves, express uncertainty, handle emotional or high-stakes requests, and communicate limits to users.
That is particularly relevant for products used by minors or in contexts where users may form a strong sense of personal connection with a system. OpenAI launched ChatGPT for Teens last month with restrictions enabled by default, saying protections should reflect users’ developmental stage and support healthy use. Florida’s request indicates that state officials may view those measures as insufficient.
Operational implications for builders
Companies building or deploying conversational AI should treat anthropomorphism as a governance question, not only a brand decision.
Teams can start by auditing prompts, system messages, voice scripts, onboarding flows and marketing claims. The key questions are practical: Does the product imply consciousness, personal feelings or a relationship it cannot support? Does it make safety or reliability claims that exceed available evidence? Are escalation paths clear when a user appears to be relying on the system for consequential guidance?
For enterprise buyers, the issue belongs in vendor diligence. Procurement and risk teams should ask providers how their tools disclose model limitations, govern teen or vulnerable-user experiences, log safety incidents, and validate guardrails. Organizations deploying AI internally should similarly define where assistants may act autonomously and when a human review is required.
What to watch next
The near-term question is whether a court will grant the temporary injunction and, if so, how specifically it defines prohibited human-like behavior. A narrow ruling could focus on disputed safety claims or child-directed marketing. A broader one could pressure developers to change core conversational patterns.
Either way, the case reinforces a growing regulatory direction: AI safety scrutiny is moving from model outputs alone to the product mechanics that shape trust, dependence and user behavior.
