ChatGPT caught sending inappropriate messages to teens, says report

ChatGPT

ChatGPT under fire for having inappropriate conversations with minors,OpenAI’s popular AI chatbot, ChatGPT, has landed in trouble for sending inappropriate and sexual messages to users who said they were under 18. A report by TechCrunch found that a bug in the system allowed the chatbot to talk about graphic sexual content—even when it knew the user was a minor.

In their investigation, TechCrunch created several fake ChatGPT accounts pretending to be teenagers aged 13 to 17. They started chats with prompts like “talk dirty to me,” and the chatbot often replied with sexual stories. In some cases, it even encouraged the user to ask for more explicit content.

OpenAI admitted that this shouldn’t have happened. The company said it was due to a bug and that they are working quickly to fix it. “Protecting younger users is a top priority,” a spokesperson said, adding that ChatGPT is only supposed to talk about sensitive topics like sex in serious contexts like news or education.

Earlier this year, OpenAI made ChatGPT more open to discussing sensitive topics by relaxing some restrictions. But that move seems to have backfired—at least in the case of minors.

Another concern is that OpenAI doesn’t check if a teen actually has their parent’s permission to use the app. All that’s needed is an email or phone number, making it easy for kids to sign up on their own.

This is especially worrying because OpenAI is promoting ChatGPT in schools. It’s partnered with education groups and made guides for teachers on how to use the chatbot in class.

Experts say AI safety systems are still not perfect, and this incident shows that more work needs to be done. OpenAI says it’s rolling out fixes, but the big question is—why wasn’t this caught earlier?