AI Conversations May Be Monitored and Shared with Authorities in Cases of Threats or Investigations

The recent case of Mally and Liel Yehalomi, who disappeared for several days and sparked a police investigation, highlights concerns about privacy in AI conversations. Investigators were able to trace some of their interactions with Google's Gemini AI, despite some data being deleted, as the sisters asked about ways to disappear and avoid detection. This case raises questions about how private AI chats really are and under what circumstances information might be shared with law enforcement.

Millions of users engage daily with AI tools like GPT, Claude, Copilot, and others, often sharing sensitive personal information. These AI services have safety mechanisms, usage policies, and protocols for sharing data with authorities. For example, OpenAI states that if its systems detect a user planning harm to others, the conversation may be flagged for human review. If a credible, immediate threat is confirmed, OpenAI may report the case to law enforcement.

Specific language indicating a concrete threat, including details about targets, timing, or methods, triggers these safety protocols. Other AI providers such as Anthropic (Claude), Google (Gemini), and Microsoft (Copilot) also maintain systems to detect and prevent dangerous or illegal use. In some cases, designated staff may access conversations for compliance checks.

Additionally, AI companies comply with official law enforcement requests for user data when legally required. They publish transparency reports and have procedures for urgent requests involving risks of death or serious injury. Users should be aware that AI chats, while seemingly private, occur within systems that store and process data, enforce safety rules, and adhere to legal obligations.