AI Assistant Exposes Mother and Daughter Trying to Disappear, Highlighting Privacy Risks

The case of Mally and Liel Yehalomi, a mother and daughter who attempted to disappear in South America, has raised significant concerns about privacy risks associated with AI chatbots. It is suspected that the chatbot Gemini may have inadvertently revealed their plans to authorities, though details remain unclear. It is not known how the police accessed their conversations, whether through an open Google account, family consent, or a court order, and what role this information played in locating them.

Unlike traditional searches or messaging apps, AI chatbots like Gemini engage users in conversational interactions that can reveal deeper personal insights. Users often share fears, plans, and decision-making processes, which the AI records and can potentially use as evidence in investigations. The Israeli Privacy Protection Authority has warned that prompts given to AI may contain sensitive personal data, urging users to minimize identifiable details.

The data collected by chatbots can build comprehensive profiles beyond typical personal information, including preferences, relationships, and thought processes. This data may be retained for extended periods, Gemini stores activity by default for 18 months, and can be used by technology companies for AI training or shared with law enforcement under legal conditions. Deleting chat history does not guarantee immediate erasure from servers.

Experts advise that while AI chatbots are valuable tools, users should not treat them as confidential confidants. It is important to review privacy settings, understand data retention policies, and consider whether shared information is appropriate for storage on commercial servers. The Yehalomi case serves as a cautionary example that technology can assist in many ways, but disappearing without a trace is not one of them.

Dr. Tehila Schwartz Altshuler, head of the Democracy in the Information Age program at the Israeli Democracy Institute, authored this analysis, emphasizing the need for awareness about AI privacy implications.