AI Chatbot Logs Become Evidence in Israeli Criminal Cases

AI conversations are becoming courtroom evidence

Israeli investigators have begun using AI chatbot conversations as evidence in criminal cases, according to reporting by Calcalistech. Two recent investigations demonstrate how interactions with systems like ChatGPT and Google Gemini can preserve digital records of user intentions, plans, and movements—often long after users assume the information has been deleted.

In one case, an 18-year-old suspect allegedly consulted ChatGPT before setting fire to a restaurant. According to the indictment, the suspect asked the chatbot what punishment could be imposed for burning down a store. When ChatGPT indicated the offense could carry a lengthy prison sentence, the suspect followed up with a second question: what would happen if the premises were closed and no one was present.

That second query proved particularly significant to investigators. Rather than showing mere curiosity about legal consequences, it suggested the suspect was actively considering how the presence or absence of people would affect the severity of charges—a potential window into premeditation.

In a separate case involving a mother and daughter who disappeared from Israel, conversations with Google's Gemini reportedly helped investigators track their movements. According to sources familiar with the investigation, the Gemini account contained questions about purchasing phones without identification, minimizing digital traces, and which South American countries lack extradition treaties with Israel. The chatbot effectively functioned as a digital diary of the planning process.

Why it matters

As AI assistants become more integrated into daily life, people increasingly treat them as private confidants—asking questions they might never pose to another person. But these conversations generate data that can be preserved, accessed through legal processes, and reconstructed into timelines of intent. For enterprise leaders, this creates both compliance obligations and employee education challenges around what constitutes appropriate use of AI tools, particularly when discussing sensitive business matters or personal plans.

The privacy paradox

The intimacy of chatbot interactions creates a false sense of privacy. Users formulate detailed queries, explain personal circumstances, and reveal information they would hesitate to share elsewhere. Yet these conversations occur within digital services that generate and store data, often in ways users don't fully understand.

Deleting a conversation from a user interface doesn't necessarily eliminate the underlying data. Information may remain accessible on devices, in cloud storage, or through forensic recovery tools. AI companies may also retain data for security, legal, or operational purposes even after users believe they've erased it.

Major AI platforms offer various privacy controls. ChatGPT allows use without account login and offers a Temporary Chat mode. Gemini provides similar temporary chat options and activity settings. Anthropic's Claude includes an incognito mode. But none of these measures guarantee absolute privacy, and some features may require sacrificing access to conversation history.

What organizations should know

The most effective privacy measure remains the simplest: don't tell an AI system anything you wouldn't want documented. Users should avoid entering identifying information unnecessarily and frame sensitive questions as generally as possible. The more specific the information provided, the more useful that conversation becomes to anyone reconstructing events.

For investigators, AI chatbots represent a new category of digital evidence—one that captures not just actions but the thought processes behind them. The technology didn't play detective in these cases; it became a witness, preserving records of what users were thinking, planning, and trying to accomplish.

As Calcalistech first reported, these Israeli cases signal a broader shift in how digital evidence is collected and interpreted. The same AI systems designed to provide helpful, personalized responses are simultaneously creating detailed records of user intentions that can persist far longer than the conversations themselves.

Details of both investigations were originally reported by Elihay Vidal for Calcalistech.