In the legal fight between Elon Musk and OpenAI, something unexpected surfaced: OpenAI president Greg Brockman's personal diary was forced into evidence. As Axios reported today, that kind of privacy invasion is spreading fast into AI territory. Courts are increasingly accepting transcripts of users' conversations with chatbots as admissible evidence—and those chats don't come with the same protections as a conversation with a lawyer or therapist.
This isn't a hypothetical. In February, a federal judge ruled that a man's conversation with Claude, where he rehearsed for a lawyer meeting, could be used as evidence against him in a criminal case. Then there's a murder case in Florida and an arson case connected to the LA wildfires—both featuring chatbot logs as courtroom exhibits.
Legal experts warn that AI chat records can be even more damaging than a personal diary. A diary is a one-way record, but large language models are designed to keep you talking, to ask follow-up questions that draw out details you might never have written down. Add precise timestamps, and the model can pinpoint exactly what you knew and when you knew it. If courts don't establish some form of immunity for AI conversations, the risk of having your digital confessions subpoenaed could reshape litigation.