OpenAI to Tighten ChatGPT on Violence and Limit Weapons Information for Teens
IA en un minuto newsroom · Editor: Jon Elgezabal
In 30 seconds
Under the new limits, ChatGPT will be terser when discussing earlier massacres: it will not piece together the weaponry used or where it came from, no matter who asks. For teenage users, the assistant will also restrict practical details about guns and security systems. OpenAI will be able to interrupt dialogues obsessed with violence and suspend the account for a while if the pattern continues.
OpenAI is adding new limits to ChatGPT, its AI assistant, for conversations that touch on violence.
When asked about mass attacks from the past, ChatGPT will give less detail and will not reconstruct which weapons were used or how the attackers got them, whatever the age of the person asking. Users under 18 will face extra limits on practical information about weapons and security systems, even if the question looks harmless. If a conversation shows a fixation on violence or on seriously hurting someone, ChatGPT will be able to cut it off, and if that keeps happening, OpenAI may temporarily pause the account.
The company will also release a community safety report on preventing violence every six months, the first due in early 2027. According to OpenAI, it sent more than 200 referrals to law enforcement in the third quarter of 2026. The changes will start arriving over the next few weeks.
Why it matters · analysis and opinion
The change does not depend on age alone: part of the restrictions will reach anyone who asks about attacks that have already happened. A company that uses ChatGPT to research security topics, write reports or train its staff may get terser answers on those subjects, and it is worth knowing that before the measures arrive. There is also an operational risk: if the system reads a conversation as a fixation on violence, the account can be paused, and what has been published does not explain how that will be judged or whether there will be a way to appeal. In the meantime, it is worth reviewing which internal uses touch on these matters, avoiding having all activity depend on a single account and following the company's twice-yearly reports.
Official source: OpenAI · Written with the help of AI: how we make the news
