AI chat histories are increasingly used as courtroom evidence, experts warn

Court filings reveal that AI chatbot transcripts have been introduced as evidence in at least twelve recent cases, leading to several convictions. The incidents span from a trial over alleged social-media addiction to a Brazilian man who unsuccessfully sought hit-man advice from ChatGPT. Since chat logs are retained on the providers' data centers and frequently remain on users' phones, they can be obtained through discovery or subpoenas regardless of a person's status as a suspect.

The lack of privileged communication protections means these records are fully accessible to authorities. To mitigate risk, analysts suggest either refraining from AI use, opting for privacy-oriented platforms such as DuckDuckGo's duck.ai, Proton's Lumo, or Brave's Leo, or installing a local large language model that never leaves the device. These steps aim to keep personal thoughts from being used against individuals in legal proceedings.