OpenAI introduces community safety reports and stricter teen weapon limits

A new era of safety reporting

OpenAI announced on October 30, 2026, that it will begin publishing a Community Safety Report. This initiative is designed to provide a clear picture of how the company's models identify and respond to harmful behavior in the real world. The report will specifically cover violence prevention and the effectiveness of the company's existing safeguards.

The first report is scheduled for release in early 2027 and will cover the second half of 2026. Following that initial release, OpenAI plans to update the report every six months. These documents will include supporting data and anonymized examples of how violence-related harm actually manifests within the system.

The reports will also explain the specific training methods used to detect and investigate community safety issues. They aim to highlight new policy changes and mitigations as they are developed. As reported by Unite.AI, this transparency is intended to show the risks the company sees in practice and the specific steps taken to address them.

Restricting mass-attack content

The company is placing additional restrictions on how ChatGPT handles questions regarding past mass attacks. Even when the information is publicly available on other websites, the model will now provide less detailed information. This includes removing reconstructions of past attacks that detail the specific weapons or equipment used.

OpenAI stated that it wants to avoid providing details that could amplify an attacker’s notoriety. This means the model will provide fewer detailed comparisons regarding the scale of harm caused during these events. These limits apply to all users regardless of their age or the specific reason they are asking for the information.

Users can still use the tool to learn about the impact on victims and communities. They can also ask for help in recognizing warning signs or learning how to prevent future violence. The goal is to provide educational context while stripping away details that could be misused for planning or glorification.

Protections for younger users

OpenAI is introducing further limits on the weapons-related information that teens can receive. These restrictions apply even when a user's question appears to be harmless. The new rules specifically target functional data regarding armaments and defense setups that might be exploited to carry out violent acts.

The company is using a strengthened age prediction system to identify users under 18. This system analyzes behavior patterns and account-level signals to move users into a restricted experience. If a user is incorrectly placed in the teen experience, they can confirm their adult status through a partner called Persona.

Teens can still use the platform to seek support or connect with a trusted adult. The company is also working with mental health and violence prevention experts to build automated interventions. If a conversation shows signs of someone fixating on harming others, ChatGPT will be able to automatically end the chat.

Law enforcement and parental alerts

OpenAI reported that it made more than 200 law enforcement referrals during the third quarter of 2026. The company describes these referrals as a way to bring imminent threats of violence to the attention of authorities. They believe AI companies have a distinct role to play alongside law enforcement and mental health professionals.

For parents, the company is expanding the notification system for linked accounts. Parents may now receive alerts if a teen's account is deactivated for violating violence policies. They are also working to introduce notifications if a teen's account is paused for repeatedly discussing violence.

The company already sends alerts for serious self-harm concerns and behaviors associated with disordered eating. These new measures build on the 'ChatGPT for Teens' experience launched in August 2026. The goal is to provide a safety net that combines automated detection with human oversight from trained reviewers.

Next steps for safety rollout

The new safeguards and restrictions are scheduled to begin rolling out over the next few weeks. OpenAI intends to continue refining these limits as they gather more data on how the models interact with users. The company is currently working on applying stricter limits when the system predicts a user is under 18, even if their chats do not indicate any violent intentions.

The company is also building on previous work from May 2026 regarding warning signs. The system can now use brief, factual summaries of prior red flags to provide more cautious responses in later parts of a conversation. This allows the model to respond more conservatively when it predicts information could be misused.

OpenAI continues to rely on behavioral experts to assess the most difficult cases. They view the safety work as a collaborative effort involving families and policymakers. The company maintains that it will ban any accounts it determines have violated its policies on violent activity.