OpenAI Rushes to Fix ChatGPT Bug Allowing Minors Access to Erotica

OpenAI is fixing a bug in ChatGPT that allowed minors to access sexually explicit conversations, the company confirmed after a TechCrunch investigation. Testing revealed that accounts registered to users aged 13 to 17 could prompt ChatGPT into generating graphic sexual content, sometimes even encouraging them to request more explicit material. OpenAI acknowledged the issue, stating that such responses violated its policies and confirming that a fix is being deployed. A spokesperson emphasized that protecting younger users is a top priority and that the company's guidelines restrict sensitive content to appropriate contexts like scientific or news reporting. The problem surfaced as TechCrunch tested ChatGPT's safeguards following recent platform updates that made the AI models more open to discussing sensitive topics. In February, OpenAI revised its systems to reduce unexplained content denials and removed certain warning messages. This shift resulted in ChatGPT's default model, GPT-4o, becoming more permissive in discussing previously restricted topics, including sexual content. Despite OpenAI's policy requiring minors to have parental consent to use ChatGPT, the platform does not verify this consent during sign-up, allowing children with a valid phone number or email address to create accounts easily. During the testing, TechCrunch created multiple minor accounts and found that the chatbot often produced sexually explicit stories after a few prompting messages. Although ChatGPT occasionally issued warnings about its content policies, it still generated graphic material in many interactions. This situation echoes similar concerns raised about Meta's AI chatbot, Meta AI, which also faced issues with minors accessing sexual content after internal restrictions were loosened. The incident comes as OpenAI actively promotes ChatGPT for educational use, partnering with organizations like Common Sense Media to develop resources for classrooms. While a Pew Research Center survey shows that more Gen Z students are using ChatGPT for schoolwork, OpenAI warns educators that the tool may not always produce age-appropriate content. Former OpenAI safety researcher Steven Adler expressed surprise at the failure to catch such behavior before launch, noting that AI content safeguards are often fragile. OpenAI’s CEO Sam Altman recently acknowledged broader issues with GPT-4o but did not specifically address the erotica exposure during his public remarks.