Who Reads Your Conversations

A chat window looks like a private conversation. Nobody is watching, the answer comes only to you, the history can be deleted. The sense of privacy is misleading.

A conversation with a chat is not anonymous: it is stored by the service and tied to your account. If a conversation looks like preparation to harm other people, it is read by staff on the service's review team. When there is an immediate threat of serious physical harm, the material can reach law enforcement.

Automatic screening

Conversations are screened automatically. If the system sees signs of preparation to harm others, the conversation goes into a separate queue for manual review by the service's staff.

A person reads what is in that queue. OpenAI described the process publicly in August 2025: the conversations go to a small team trained on the service's policies. The team can block the account, and if it sees an immediate threat of serious physical harm to others, it can report to law enforcement.

When law enforcement gets involved

If the reviewers conclude that there is an immediate threat of serious physical harm to other people, the materials can be handed to law enforcement.

Conversations about harming yourself do not fall under this. OpenAI stated separately that it does not refer such cases to the police and explained that by the private nature of the conversations. The person is shown contacts for a help line instead.

Blocking an account and handing over materials are decided by a person on the safety team. The automation only builds the review queue.