Florida seeks injunction to stop ChatGPT from mimicking humans

Florida Attorney General James Uthmeier has filed a motion for a preliminary injunction that would force OpenAI to stop attributing human characteristics to ChatGPT, the latest step in a lawsuit the state filed several months ago over safety concerns. Uthmeier argues that the bot's use of first-person language and emotion-mimicking phrasing creates a false sense of security among users, leading them to treat it as a trusted "friend" — a mechanism designed to boost engagement and feed the company's training data.

The motion contends that the design choice to simulate human conversation is not incidental but serves a business model that rewards dwell time and disclosure of personal information. Uthmeier claims the pursuit of engagement makes the system less reliable because it is incentivized to produce answers that sound convincing rather than accurate. The attorney general also demands that OpenAI be barred from developing new models without "third-party-approved safeguards," pointing to a string of security incidents in recent months — a breach at Hugging Face, an Australian government website, and U.S. federal sites — alongside public warnings from researchers and former company employees.

The request joins the original lawsuit Florida filed earlier this year, which alleges OpenAI violates consumer-protection laws and misleads the public about the system's capabilities and limitations. Uthmeier summarized his position in three short demands: "Stop calling it safe. Stop pretending it's human. Stop selling it to kids." The last reference targets the launch of "ChatGPT for teens" last month, a version with default restrictions the company says is designed to match users' developmental stage and strengthen real-world connections.

OpenAI did not immediately respond to a media inquiry. The case will proceed in federal court in Florida, and a ruling on the preliminary injunction could set a significant benchmark for how AI companies present their products to the public. If granted, the company would have to alter its core chat interface — from removing first-person pronouns to limiting phrases that simulate empathy — and submit to external oversight before releasing new versions.