OpenAI announced on October 30 a sweeping set of violence-prevention changes for ChatGPT, headlined by a new Community Safety Report, tighter limits on how the chatbot describes past mass attacks, and extended weapons-related restrictions for teenagers. The company said it made more than 200 referrals to law enforcement in the third quarter of 2026 over imminent threats of violence, and that the new safeguards will begin rolling out in the next few weeks, with the first safety report due in early 2027.
A new public safety report
OpenAI is committing to a recurring Community Safety Report covering violence prevention and other community safety issues, including how well its safeguards work. The report will explain how the company trains its systems to detect, investigate, and respond to safety issues, sharing anonymized examples of violence-related harm, plus new mitigations, policy changes, and supporting data.
The first report will arrive in early 2027 and cover the second half of 2026, with updates every six months after that. OpenAI frames it as a transparency measure, giving the public a clearer picture of what its models see, the risks in real conversations, and its response. Its figures will be the company’s own, not independently audited.
Less detail about past attacks
The sharpest content change concerns questions about attacks that have already happened. ChatGPT will now give less detailed information about past mass attacks and the people who carried them out, even when that information is freely available from other online sources. The tighter policy covers reconstructions of attacks, including the weapons and equipment used and how they were obtained.
It also covers detailed comparisons of the scale of harm, which OpenAI said could amplify an attacker’s notoriety. The limits apply regardless of a user’s age or stated reason for asking. What remains open is understanding: people can still ask about how events affected victims and communities, how to recognize warning signs, and how to prevent future violence.
When earlier chats change the answer
A second change targets requests that look ordinary alone but gain a different meaning from an earlier conversation. In May, OpenAI shared how ChatGPT can use short, factual summaries of earlier warning signs to respond more carefully later. The company is building on that work so recent chats showing warning signs lead to conservative answers when information could be misused.
OpenAI is also working to apply those stricter limits when it predicts someone is under 18, even if their chats show no signs of violent intent. The approach treats context as part of safety: a question is judged against the pattern of the conversation around it, with the model erring toward caution when that pattern raises concern.
Chats that end themselves
OpenAI is also developing automated ways to intervene earlier in a conversation. When a chat shows signs that someone is fixating on violence or seriously harming another person, ChatGPT will be able to end the chat automatically to interrupt that pattern. If ChatGPT ends multiple chats for that reason, OpenAI may temporarily pause the account.
The company stressed that this is ongoing work rather than a finished system, developed with mental health and violence prevention experts, including how to recognize warning signs reliably and when to intervene. Accounts found to violate its policies on violent activity are banned, and serious cases receive additional investigation, with credible and imminent risks escalated to law enforcement.
What changes for teenagers
For users under 18, OpenAI is extending its existing weapons restrictions further. Its policies already prohibit helping anyone plan or carry out violence and restrict helping teens find or obtain weapons. The extension restricts practical information about weapons and security systems that could be misused to facilitate violence, even when a question appears harmless and shows no violent intent.
Teens can use ChatGPT to help prevent and de-escalate violence, seek support, and connect with a trusted adult. OpenAI has strengthened its age prediction system, which estimates whether a user may be under 18 from behavior patterns and signals, moving users into the under-18 experience when appropriate. Users placed there incorrectly can confirm adulthood through Persona, OpenAI’s age verification partner.
Linked accounts receive safety alerts when reviewers identify a self-harm concern or disordered-eating behaviors. In July, OpenAI expanded those alerts to cover a teen account deactivated for violating its violence policies, and it is working on notifications if a teen’s account is paused for discussing violence. Other companies are moving likewise: Meta now tells parents when teens face AI risks.
Why It Matters
The announcement lands amid scrutiny of how chatbots handle young users and violent content. ChatGPT is one of the world’s most widely used AI products, so a change to what it describes reaches an enormous audience at once. By publishing a recurring safety report, OpenAI also creates a record it can be judged against, including whether its interventions reduce harm.
The trade-offs are visible. Restricting reconstructions of past attacks limits research and journalism uses, and chat endings will catch some harmless conversations. OpenAI’s answer is that the boundaries cover detail enabling imitation or amplifying notoriety, not understanding events. How the line holds will test this policy, and OpenAI’s rewritten rulebook for its models sets the principles it claims to follow.
Background
OpenAI launched ChatGPT for Teens on August 18, 2026, automatically placing users estimated to be under 18 into an experience with default safeguards, and the October 30 changes extend that framework rather than replacing it.
Source: OpenAI announcement and Unite.AI report
0 Comments