Community Safety Report and data disclosure
OpenAI said it will publish a Community Safety Report focused on violence prevention and other safety issues. The first report will appear in early 2027 and cover the second half of 2026, with updates every six months thereafter.
The company described the report as a summary of how its systems detect, investigate and respond to potential harm. OpenAI said the document will include anonymized examples, descriptions of new mitigations and policy changes, and relevant supporting data. The company reported making over 200 law enforcement referrals in Q3 2026 and stated those referrals have helped disrupt violent plots.
Stricter limits on violence-related content
OpenAI said its models are trained to avoid providing instructions that would help plan or carry out violence. The company announced additional restrictions on responses about past mass attacks and the people who carried them out, directing models to provide less detailed reconstructions even when such details exist online.
New limits will curb descriptions of weapons, equipment, how they were obtained, and detailed comparisons of scale that could amplify an attacker’s notoriety. OpenAI said these constraints apply to all users, regardless of age or stated purpose.
Building on earlier work from May, the company said ChatGPT can use short factual summaries of prior warning signs to respond more cautiously in subsequent chats. When recent conversation history shows warning signs of possible violence, ChatGPT will adopt more conservative responses for information that could be misused.
Protections for teens, age prediction and parental controls
OpenAI described an expanded set of safeguards for ChatGPT for Teens. The company said it will place further limits on weapons-related information available to users under 18, extending restrictions to practical details about weapons and security systems that could be misused even if no violent intent is apparent.
OpenAI said it strengthened its age prediction system and continues to use Persona as an age verification partner; users placed incorrectly in the under‑18 experience will have a privacy-protective way to confirm adulthood. Parental controls allow parents to link accounts and receive safety alerts; in July, alerts were expanded to notify parents if a teen’s account is deactivated for violating violence policies. OpenAI also said it is working to add notifications when an account is paused for repeated discussions of violence.
The company said accounts that violate violent activity policies will be banned and that repeated automatic chat closures for fixation on violence may lead to temporary account pauses. OpenAI stated the new safeguards will begin rolling out in the next few weeks.
Original source: OpenAI News