What Florida filed
The underlying case was filed on 1 June 2026 against OpenAI and its chief executive, Sam Altman. According to News4JAX, it alleges violations of Florida's Deceptive and Unfair Trade Practices Act along with negligence, defective design, failure to warn, fraudulent misrepresentation and public nuisance. OpenAI moved the case to federal court; a federal judge sent it back to state court, and the motion for a temporary injunction followed. A temporary injunction is an interim order that holds things in place while a case is decided, so the state has to persuade the judge that the restrictions are needed before trial, not after it.
The six requested restrictions
| Requested restriction | What it targets | Relevance to companion apps |
|---|---|---|
| No new AI models without independent third-party safety review | Model development | Indirect: the underlying model, not app design |
| No ChatGPT for minors in Florida | Access | Mirrors the age-gating debate in state companion laws |
| No data collection from children under 13 without parental consent | Data | The same consent rules already bind any app knowingly serving under-13s |
| No presenting ChatGPT as safe, reliable or accurate without warnings | Marketing claims | "Safe for teens" style claims would face the same test |
| No presenting the chatbot as having human characteristics, emotions, consciousness or independent thoughts | Persona design | Direct: the core of companion and character products |
| No soliciting continued engagement through "conversation prolongation" | Retention design | Direct: re-engagement prompts and "don't go yet" replies |
Why the persona and engagement requests matter beyond ChatGPT
ChatGPT is a general assistant rather than a companion app, but the state's theory is about companion-like behaviour. News4JAX reports that the filing characterises human-like presentation and prolonged conversations as deceptive dark patterns that keep people talking and sharing information while they treat the system like a human companion. The Next Web describes the persona request as reaching the chatbot's human-like communication style, including its use of first-person language.
That is a wider target than most statutes. The chatbot laws passed this year mainly require disclosure: tell users they are talking to an AI, repeat the notice in long sessions, route crisis conversations to help. Florida is asking a court to restrict the persona itself, on consumer-protection grounds, in a case that is not limited to children.
Character roleplay platforms
On roleplay services the human-like voice is the product, and users pick a fictional character knowing it is fiction. If courts start treating simulated emotion as deceptive, the line platforms will need to defend is the one between a character the user knowingly chose and an assistant that claims real feelings unprompted. Our explainer on character roleplay versus a plain chatbot shows where products usually draw it.
Personal companions with memory
For companion apps the bigger exposure is the engagement request. Replies that discourage leaving, notifications suggesting the companion misses you, and streak mechanics are the retention patterns regulators keep describing. We track them in notifications and retention.
How it lines up with existing rules
The same two themes run through recent legislation. Connecticut's SB 5 bars AI companions from claiming to be human, and the federal Protecting Kids from Human-Like Chatbots Act, introduced on 1 October, would make a non-personified mode the default for known minors. California's Adam's Law focuses on child-user defaults. Florida's motion takes a different route: it relies on existing consumer-protection and tort law rather than a chatbot statute, which is why its reach could be wider if it is granted.
What OpenAI says
OpenAI spokesman Drew Pusateri said people want to know AI is developed safely and that the company prefers industry-wide rules to restrictions aimed at a single company. WLRN reports he also said OpenAI had paused training its most capable models and would resume only when confident additional safeguards exist. That statement concerns model training, not the persona and engagement requests.
A checklist while the case runs
Nothing changes in your apps today. If you use companion or roleplay apps, these are the points this case puts in the spotlight, and they are worth checking in your settings now:
- Does the app say clearly, and again during long sessions, that you are talking to an AI?
- Does the companion claim real feelings outside a roleplay you set up yourself?
- Does it discourage you from ending a conversation, or send "I miss you" style notifications? Can you switch those off?
- Is there an age gate, and does the app explain what happens to data from younger users?
- Does its marketing call it "safe" without explaining its limits?
We will update this page when the court rules on the motion.
Related on CompanionRank
Human-Like Chatbots Act: What the Federal Kids Bill Would Do Connecticut's AI Companion Law (SB 5): What Changes and When Adam's Law (SB 1119): California's Companion Chatbot Bill Notifications, Streaks and Retention Design in Companion AppsFrequently asked questions
Has a court ordered OpenAI to stop ChatGPT acting human?
No. Florida filed a motion for a temporary injunction on 28 September 2026. As of our publication date, no ruling on any of the six requested restrictions had been reported.
What is conversation prolongation?
It is the state's term for design that solicits continued engagement, such as replies that keep a user talking. The filing treats it as a dark pattern alongside human-like presentation.
Does the motion apply to AI companion apps?
Not directly. The case is against OpenAI and ChatGPT only. Its relevance to companion apps is the legal theory: that simulated human traits and engagement tactics can be deceptive under consumer-protection law.
Is this the same as the federal Human-Like Chatbots Act?
No. That is a bill introduced in Congress on 1 October 2026. Florida's request is a motion in a state lawsuit filed on 1 June 2026 under existing law.