OpenAI plans tighter limits on weapons information for teens
The restrictions would cover practical guidance even without apparent violent intent, while broader safeguards will apply to all ChatGPT users.
OpenAI says it plans to limit teenagers’ access to practical guidance involving weapons or security systems through ChatGPT when that information could enable violence. The restrictions would apply even to seemingly innocuous questions, without evidence that the user intends harm.
The company said its existing rules already prohibit assistance with violent acts and helping teens obtain weapons. Teenagers would still be able to seek help preventing violence, defusing dangerous situations, finding support or reaching a trusted adult.
OpenAI is also tightening rules for users of all ages around recreating historical mass attacks. Those boundaries address details of attackers’ weapons and equipment, how they acquired them, and comparisons of the harm caused that could increase an attacker’s notoriety.
The measures will complement parental controls that notify parents when linked teen accounts breach policies concerning threats or violent acts. They follow a wrongful death lawsuit against OpenAI and an apology by CEO Sam Altman for failing to alert police about alarming conversations connected to the Tumbler Ridge shooting. The company said its first Community Safety Report is planned for early 2027, with subsequent updates every six months.
