OpenAI reports 200+ law enforcement referrals and plans its first Community Safety Report for early 2027.
OpenAI has announced stronger ChatGPT protections for teenagers and plans to automatically end conversations showing signs of fixation on serious violence. Announced on October 30, 2026, the changes include tighter content restrictions, enhanced age prediction, and potential account pauses following repeated high-risk interactions.
The company says the new safeguards will begin rolling out over the coming weeks, while automatic chat termination remains under development.
ChatGPT Could Automatically End High-Risk Conversations
OpenAI is developing a system that could interrupt conversations involving warning signs of violence or intentions to seriously harm others. Repeated interruptions may trigger temporary account restrictions.
The approach builds on technology introduced in May 2026 that uses brief summaries of earlier safety-related conversations to recognize concerning patterns and respond more cautiously.
OpenAI has not disclosed when automatic chat termination will become widely available.
OpenAI Strengthens Teen Safety Protections
For users under 18, ChatGPT will apply tighter restrictions on practical weapons-related and security-system information, even when requests appear harmless.
OpenAI is also strengthening its age-prediction system to continuously assess account signals and identify users who may require additional teen safeguards.
New parental notifications are planned for linked accounts when repeated violence-related discussions result in temporary account pauses.
Separately, users of all ages will face stricter limits on detailed reconstructions of historical mass attacks, while educational and prevention-related discussions remain available.
OpenAI Reports 200+ Law Enforcement Referrals
OpenAI reported making more than 200 law enforcement referrals during Q3 2026, reflecting its existing efforts to identify and escalate potentially serious threats.
The figure represents referrals, not confirmed violent plots or prevented attacks.
Independent testing has also raised questions about teen safeguards. In October, Common Sense Media identified weaknesses in parental alerts and age prediction after testing more than 4,000 prompts. OpenAI disputed aspects of the assessment, which preceded the latest announcement.
First Community Safety Report Planned for 2027
OpenAI plans to publish its first Community Safety Report in early 2027, covering July–December 2026, with updates every six months.
The reports will examine safety detection, enforcement, and improvements, providing further evidence for assessing how effectively ChatGPT protects users and responds to emerging risks.
