OpenAI lets ChatGPT end chats showing violent warning signs

The latest:

ChatGPT will stop reconstructing past mass attacks — including the weapons and equipment used and how they were obtained — regardless of a user’s age or stated reason for asking. OpenAI announced the violence-related restrictions on October 30, 2026, alongside a recurring Community Safety Report and added protections for users under 18. The company said the safeguards begin rolling out in the coming weeks.

Details:

• What stays allowed: According to OpenAI, users will still be able to ask about mass attacks in order to understand what happened, the impact of such events, and the warning signs that preceded them. The restriction targets operational reconstruction — the specific weapons, equipment and acquisition methods — rather than the historical record itself.
• The trigger mechanism: OpenAI said that when a user’s recent conversations show indicators of potential violence, ChatGPT will answer more conservatively on information that could be misused. The threshold is based on signals across recent chats rather than a single message, according to the company.
• Ending the conversation: When a conversation shows fixation on violence or on harming another person, ChatGPT will be able to terminate the exchange automatically, the company said. Repeated terminations may lead OpenAI to pause the account temporarily. The company did not set out how many endings trigger a pause.
• Teen restrictions: For users under 18, OpenAI is adding limits on practical information about weapons and security systems even when no indicator of violent intent is present. That makes the teen standard stricter than the general one, which applies the broader conservatism only after warning signs appear.
• Age prediction: The company said it strengthened its age-estimation system to continuously evaluate new signals. Adults wrongly routed into the under-18 experience can confirm their age through Persona, OpenAI’s verification partner.
• Parental notification: OpenAI said it is working on alerting parents with linked accounts if a teen’s account is paused for repeatedly discussing violence. The company described the feature as in development and did not give a launch date.
• The referral figure: OpenAI said it made more than 200 referrals to law enforcement in the third quarter of 2026. The announcement did not break the figure down by country, by case type, or by outcome.
• The reporting cadence: The first Community Safety Report is due in early 2027 and will cover the second half of 2026, the company said, with updates every six months thereafter — making the Q3 referral number a preview of a metric OpenAI now commits to publishing on a fixed schedule.

Between the lines:

The package shifts ChatGPT from filtering individual prompts toward assessing a user’s recent conversation history, since the conservative-response trigger and the automatic chat ending both rest on patterns rather than single messages. The temporary account pause, paired with the planned parental alert, extends enforcement beyond the conversation into the account itself.

What’s next

The safeguards begin appearing within weeks. The first Community Safety Report, covering the second half of 2026, is due in early 2027, with the next edition six months later. The parental notification feature remains undated.