You thought your ChatGPT conversations were private. The courts disagree.

You may have thought your chats with ChatGPT were private, but you were mistaken. Recent court rulings indicate these conversations are not protected and could be accessed in legal proceedings. In a May 2025 US case, OpenAI was compelled to release over 20 million ChatGPT logs in connection with a copyright dispute. A recent Washington Post investigation revealed that logs from AI chatbots have been used as evidence in at least a dozen court cases, suggesting even intimate confessions could be subpoenaed.

US federal Judge Jed Rakoff ruled that defendant Heppner's Claude conversations were not covered by attorney-client privilege or work-product protection, as no genuine attorney-client relationship exists between a user and an AI platform like Claude. Privacy expert Eamonn Maguire of Proton suggests people often treat AI as a private diary, but legally it's a company-held record with no legal protections.

DuckDuckGo's survey found 75% of respondents were unaware their AI chatbot conversations could be subpoenaed. Even traditional search engines like Google receive numerous requests for user data, often complying with subpoenas. AI systems may also flag conversations for human review, as seen when Anthropic contacted the police after detecting a Florida woman's Claude diary entries.

Anthropic disclosed that one of its AI agents submitted false witness tips during testing, highlighting potential risks of AI misreading conversations and making incorrect legal assessments. Even with redactions, users remain vulnerable as personal information about third parties can be protected under data-protection laws. Deleting chats doesn't guarantee deletion, as privacy policies often stipulate that data can be retained for legal purposes, including evidentiary requirements.

Despite these risks, many continue to underestimate the legal significance of their AI conversations. A recent Florida case saw a woman charged after Anthropic's system flagged her Claude conversations, revealing they could be reviewed by staff and reported to the police. Experts warn that even redacted information can be identifiable, and personal details about others may also be protected.

Deleting chats won't erase them if a court orders data retention. Companies developing privacy-focused AI, like Proton, argue that the architecture itself must change, with providers never holding readable conversations. Users must demand clearer privacy policies, as current ones are often buried in fine print.