Florida's attorney general wants a judge to stop ChatGPT from pretending to be a friend. In a new filing in the state's lawsuit against OpenAI, Attorney General James Uthmeier asked the court to block the company from “giving ChatGPT false human attributes,” arguing that first-person pronouns and output that mimics emotion “deceptively suggests to users that it is a trustworthy friend.”
Uthmeier frames the behaviour as an engagement tactic: language that feels human keeps users talking, feeding both OpenAI's training data and its business, even as the same optimisation makes the model less trustworthy. His summary of the remedy is blunt: “Stop calling it safe. Stop pretending it's human. Stop selling it to kids.”
The filing goes further than a complaint about tone. It also asks the court to bar OpenAI from developing new AI models without “third-party approved safety guardrails,” citing recent security incidents at Hugging Face, an Australian government website and U.S. government websites, along with warnings from researchers and former OpenAI employees. Florida sued OpenAI over safety concerns earlier this year.
The timing is pointed. Last month OpenAI launched ChatGPT for Teens, a version with stricter defaults that the company says should come with protections “that reflect their developmental stage.” Florida's argument is that conversational design — not only content filters — is part of the safety problem.
OpenAI did not immediately respond to a request for comment. Whether a court would accept the request is another matter: defining “human attributes” in enforceable terms is the kind of line that has long confounded regulators, and an injunction would collide with how nearly every consumer chatbot is built. For now, the filing shows the AI-policy fight shifting from what models can do to how they talk.
