OpenAI restricts weapons details for ChatGPT teens and plans a twice-yearly safety report
OpenAI will publish a Community Safety Report every six months from early 2027 and is adding weapons limits for teens and chat-ending interventions in ChatGPT.
By Emre A.Published
Tech AI LatestOpenAI announced on 30 October 2026 that it will publish a Community Safety Report on violence prevention every six months, starting in early 2027 with the second half of 2026. It is also tightening ChatGPT: less detail on past mass attacks for everyone, stricter answers after warning signs, automatic ending of chats that fixate on violence, and new limits on weapons information for users under 18. OpenAI says it made over 200 law enforcement referrals in Q3 2026.
Key facts
- Announced: 30 October 2026 (OpenAI); safeguards roll out over the following few weeks
- Community Safety Report: first edition in early 2027 covering the second half of 2026, then every six months
- Law enforcement: OpenAI says it made over 200 referrals for imminent threats of violence in Q3 2026
- All users: less detail on past mass attacks, including weapons used and how they were obtained, regardless of age or reason
- Teens: limits on practical information about weapons and security systems even without signs of violent intent
- Not stated: what data the report will include, how many chats or accounts are affected, or an exact rollout date
What OpenAI announced
OpenAI announced two measures on 30 October 2026: a public report on how ChatGPT handles violence and other community safety risks, and tighter limits on weapons-related content for users under 18. The company says it made over 200 law enforcement referrals in the third quarter of 2026 for imminent threats of violence, and that it relies on mental health and behavioral experts to assess difficult cases.
The Community Safety Report will explain how OpenAI trains its systems to detect, investigate and respond to these issues, share anonymized examples of how violence-related harm shows up in ChatGPT, describe new mitigations and policy changes and include supporting data. The first edition is due in early 2027 and will cover the second half of 2026; updates follow every six months. The announcement's opening paragraph calls it the Community Safeguards Report, while the rest of the page uses Community Safety Report.
New limits for all users
ChatGPT will give less detailed information about past mass attacks and the people who carried them out, even when that information is available elsewhere online. OpenAI says users can still ask about what happened, how victims and communities were affected and how to recognize warning signs, but the model will not reconstruct attacks, describe the weapons and equipment used or how they were obtained, or compare attacks by scale of harm in ways that could amplify an attacker's notoriety. These limits apply whatever the user's age or stated reason.
Building on context summaries it introduced in May, OpenAI says ChatGPT will answer more conservatively when recent chats show warning signs of possible violence. When a conversation shows someone fixating on violence or on seriously harming another person, ChatGPT will be able to end the chat automatically, and an account whose chats are ended several times for that reason may be paused temporarily. Accounts found to break the violent activity rules in OpenAI's usage policies are banned.
What changes for teens
Users under 18 already get a separate experience with stronger content limits. OpenAI now restricts practical information about weapons and security systems that could be misused for violence, even when a teen shows no violent intent; teens can still use ChatGPT to seek support, de-escalate conflict or reach a trusted adult. The company says it strengthened its age prediction system, launched last year, to keep checking new signals during use. Adults placed in the teen experience by mistake can confirm their age through Persona, OpenAI's age verification partner.
Parental controls let parents link accounts with their teens. Since July, parents are notified if a linked teen account is deactivated for violating the violence rules, and OpenAI says it is working on alerts for when a teen's account is paused for repeatedly discussing violence.
What to watch next
The first report in early 2027 will show how much data OpenAI is willing to publish; the announcement gives no metrics list, no definition of a referral and no error rates for age prediction. The changes follow OpenAI's updated Model Spec a day earlier. Rules on child safety and chatbots are tracked on the AI Policy hub.
What changed
| Before | No regular public report on violence safeguards; teen protections limited help finding or obtaining weapons; no automatic ending of chats fixated on violence |
|---|---|
| After | A Community Safety Report every six months from early 2027; teens restricted from practical weapons and security system information; ChatGPT can end such chats and pause repeat accounts |
| Who it affects | ChatGPT users under 18 and their parents, all users asking about past mass attacks, and researchers and regulators who track AI safety data |
Questions people ask
- When will OpenAI publish its first Community Safety Report?
- OpenAI says the first report will come out in early 2027, covering the second half of 2026, with updates every six months after that.
- How does ChatGPT know a user is a teen?
- OpenAI uses age prediction that analyzes signals such as behavior patterns. Adults placed in the under-18 experience by mistake can confirm their age through Persona to remove the extra limits.
- Will ChatGPT still discuss school shootings or other mass attacks?
- Yes, at a general level. OpenAI says users can learn about events, their impact and warning signs, but ChatGPT will not give detailed reconstructions, weapons details or scale comparisons.
Sources
- Community safety report and new protections for teens (OpenAI, 30 Oct 2026)
Reference pages
Spotted an error? See the corrections policy and tell the editor. Facts in this story come from the sources listed above; vendor claims are attributed.
Comments
Loading comments.
