OpenAI tightens ChatGPT violence safeguards and teen weapon limits
OpenAI has announced a Community Safety Report and new ChatGPT restrictions on firearms and violence-related content for minors. Its first report is scheduled for early 2027, covering the second half of 2026, followed by updates every six months. The company also said it made more than 200 law-enforcement referrals in the third quarter of 2026 for imminent threats of violence.
The announced safeguards will begin rolling out in the next few weeks. They cover all users in some circumstances, while adding tighter controls for people identified as under 18 through ChatGPT for Teens and OpenAI’s age-prediction systems.
A recurring public account of safety operations
OpenAI says the Community Safety Report will describe how its systems detect, investigate and respond to violence prevention and other community safety issues. It will include anonymized examples of violence-related harm, new mitigations and policy changes, and data intended to support the findings.
The company said it combines automated detection, review of concerning activity and escalation of credible threats. It also relies on mental-health and behavioral experts to assess difficult cases, while positioning AI providers as one participant alongside law enforcement, families, clinicians and policymakers.
The report is intended to make the company’s safeguards and their effectiveness more visible. Its publication cadence creates a defined channel for disclosures about harmful-content patterns, response measures and subsequent changes to protections.
Less operational detail about attacks and weapons
For all users, ChatGPT will provide less detailed information about past mass attacks and those responsible for them. OpenAI is restricting reconstructions that cover weapons, equipment and acquisition, as well as detailed comparisons of the scale of harm that could amplify an attacker’s notoriety.
Users may still seek information about the impact on victims and communities, warning signs, and violence prevention. OpenAI says the tighter boundaries apply regardless of the user’s age or stated reason for asking.
The company is also extending context-sensitive controls. When recent chats indicate possible violence, ChatGPT will be more conservative with information that could be misused. If a conversation shows signs of fixation on violence or seriously harming another person, OpenAI says the service may automatically end the chat; repeated endings may result in a temporary account pause.
Additional limits for under-18 users
For younger users, the expanded teen experience in the expanded teen experience applies age-appropriate safeguards while the new policy restricts practical information about weapons and security systems that could facilitate violence, even when a request appears harmless. Teens can still use ChatGPT for prevention and de-escalation support, and to seek help from a trusted adult.
OpenAI says its age-prediction system assesses multiple signals, including behaviour patterns, and has been strengthened to reassess signals as people use ChatGPT. People incorrectly put into the under-18 experience can confirm adulthood through Persona, OpenAI’s age-verification partner. Parents can link accounts with teens and manage safety settings; the company is also working on notifications when an account is paused for repeated discussions of violence.
For businesses deploying conversational AI, the announcement underlines the practical importance of age-sensitive controls, clear escalation processes and regular evidence-based reporting on how safety policies are applied.