OpenAI will allow erotic content in ChatGPT for users who verify their age, Sam Altman said in a post on X on Tuesday. The feature arrives in December, tied to a fuller age-gating rollout, under what Altman calls the principle of treating adult users like adults. In the same post he said a new version of ChatGPT that behaves more like what people liked about GPT-4o will ship in the coming weeks. “If you want your ChatGPT to respond in a very human-like way, or use a ton of emoji, or act like a friend, ChatGPT should do it,” he wrote, adding: “But only if you want it.”

The framing reverses months of public caution. Altman wrote that ChatGPT was made “pretty restrictive to make sure we were being careful with mental health issues,” that this made it less useful and enjoyable for users without such problems, and that the company can now safely relax restrictions in most cases. OpenAI pointed to age prediction, parental controls and a router in GPT-5 that flags concerning behavior. The company also announced an eight-member expert council on well-being and AI; Ars Technica observed it contains no suicide-prevention specialists, and OpenAI also faces scrutiny over a lawsuit filed by a California couple who alleged ChatGPT contributed to their 16-year-old son’s suicide.

The post drew immediate replies noting an August interview in which Altman cited not having put a sex bot avatar in ChatGPT as a decision made for the world rather than the race. Rival xAI previously launched flirty AI companions in Grok. What the new policy will mean for voice, image and video tools is unspecified.

The policy shift comes as OpenAI faces scrutiny over user safety even as it pushes ChatGPT toward a more permissive adult mode.