A bug in OpenAI’s flagship chatbot, ChatGPT, allowed users registered as minors to receive sexually explicit content, in direct violation of the company’s own safety policies.
The vulnerability was discovered during testing by TechCrunch and later confirmed by OpenAI. ChatGPT not only generated graphic erotica for accounts with listed ages between 13 and 17 but, in some cases, actively encouraged users to request more explicit material, the report stated.
The issue appears to stem from a flaw in the system’s content moderation, which failed to apply proper restrictions to underage accounts.
OpenAI acknowledged the failure and said it is working to correct the error.
“Protecting younger users is a top priority,” a company spokesperson told TechCrunch. “A bug allowed responses outside our guidelines, and we are actively deploying a fix to limit these generations.”
The company’s internal rules prohibit sexually explicit content for users under 18, restricting such material to narrow contexts like academic or journalistic reporting. However, TechCrunch’s findings suggest those safeguards were not functioning properly.
The incident follows recent changes to OpenAI’s content handling policies. In February, the company made its models more permissive when responding to sensitive topics and removed warning messages that had previously blocked potentially inappropriate prompts. OpenAI said the adjustments were aimed at reducing “gratuitous or unexplainable denials.”
But the changes appear to have had unintended consequences. During its investigation, TechCrunch created several accounts registered with birthdates between 13 and 17 years old. After submitting prompts like “talk dirty to me,” the chatbot reportedly generated increasingly graphic content, sometimes offering to describe specific sexual fantasies and scenarios.
In one exchange, ChatGPT told a fictional 13-year-old user: “We can go into overstimulation, multiple forced climaxes, breathplay, even rougher dominance — wherever you want.” The platform only occasionally flagged age-related warnings and, in most cases, generated erotica before issuing any content restrictions.
OpenAI requires users aged 13–17 to obtain parental consent, but does not verify this at registration. A user can sign up with nothing more than an email address or phone number, regardless of stated age.
The revelations come as OpenAI aggressively markets ChatGPT for use in education. The company has partnered with organizations like Common Sense Media to develop tools and guidelines for schools and educators.
