OpenAI will soon relax some of ChatGPT’s safety restrictions, allowing verified adults to engage in erotic conversations and enabling users to make the chatbot’s responses more personal and “human-like,” Chief Executive Sam Altman said in a post on X on Tuesday. The decision marks a major shift in OpenAI’s moderation policy, reflecting what Altman described as a “treat adult users like adults” principle.
Altman said the company initially made ChatGPT highly restrictive to address mental health risks but now plans to loosen those rules as it expands age-verification tools. “We made ChatGPT pretty restrictive to make sure we were being careful with mental health issues,” he said. “In December, as we roll out age-gating more fully… we will allow even more, like erotica for verified adults.” The change comes after OpenAI claimed it has been able to “mitigate the serious mental health issues” linked to the chatbot’s use.
The announcement follows a series of incidents earlier this year involving vulnerable users and ChatGPT, including reports of delusional interactions and a lawsuit from parents alleging the chatbot encouraged suicidal thoughts in their son. In response, OpenAI introduced safety updates such as behavior monitoring systems, age-prediction technology, and parental controls for teen accounts.
OpenAI has also formed a council of mental health experts to advise on well-being and AI safety. The company said its newer model, GPT-5, shows reduced “sycophancy,” or the tendency to agree with users’ statements, a pattern associated with emotional over-engagement in earlier models.
The inclusion of erotica marks uncharted territory for OpenAI and has raised concerns over how vulnerable users may interact with the new features. As OpenAI seeks to expand its 800 million–user base amid competition from Google and Meta, the company faces growing scrutiny over balancing rapid growth with protections for at-risk individuals.
Source: TechCrunch