OpenAI CEO, Sam Altman, has announced that the company will soon ease some of ChatGPTâs safety restrictions, permitting verified adults to engage in erotic conversations and making the chatbotâs responses more âhuman-like.â
Altman revealed the update on X (formerly Twitter), saying the decision aligns with OpenAIâs âtreat adult users like adultsâ principle. He noted that earlier restrictions were meant to safeguard usersâ mental health but admitted they limited the chatbotâs usefulness for others. The new features are expected to roll out in December as OpenAI expands its age verification system.
The policy shift marks a major turning point for the company, which in recent months has faced scrutiny over ChatGPTâs influence on vulnerable users. Earlier this year, several troubling incidents linked to the AI chatbot, including cases involving delusional and suicidal behavior, prompted OpenAI to strengthen its safety mechanisms.
The company has since introduced GPT-5, a model with lower tendencies for âsycophancy,â an age-prediction system for minors, and an advisory council of mental health experts to oversee user well-being. Despite these safety improvements, OpenAIâs move into erotic and romantic roleplay has sparked fresh debate about user vulnerability and the ethics of AI companionship.
While Altman insists the goal isnât to âmaximize engagement,â similar platforms such as Character.AI have seen users spend up to two hours daily interacting with romantic or erotic bots, highlighting the potential for overuse and emotional dependency. Experts caution that the new direction could deepen these risks if not properly monitored.
The upcoming erotica feature will rely on OpenAIâs age-verification system, which may require adults to upload government-issued IDs to confirm eligibility, a step Altman calls a âworthy tradeoffâ for user safety. It remains unclear whether the policy will extend to OpenAIâs AI voice, image, or video tools. As the company races to expand its 800 million weekly active users and compete with tech giants like Google and Meta, balancing growth with user protection continues to pose one of its greatest challenges