Be Careful with AI: Your Revelations to It May Be Used in Court

The article warns that confidential conversations with AI chatbots can be used in court. It cites the case of Bradley Hefner, whose chats with Claude became evidence in court despite lawyers' attempts to invoke attorney-client privilege. The author explains that data shared with a third party is not protected, and that AI can infer hidden information. Research shows that users often underestimate the risks, and companies make it difficult to disable data collection.

The article warns that confidential conversations with AI chatbots can be used in court. As an example, it cites the case of Bradley Hefner, CEO of an American company accused of fraud. He entered confidential information into the chatbot Claude and asked it to formulate a defense strategy. After his arrest, the FBI seized his computers and discovered the correspondence. Lawyers tried to challenge the use of these materials, citing attorney-client privilege, but a federal judge rejected the motion because the data was shared with a third party. Hefner was found guilty. The author also cites a study showing that users often underestimate the risks, and that AI can infer hidden information from conversations. It is noted that companies design interfaces to make it difficult to disable data collection. The article emphasizes that sharing information with AI is a waiver of confidentiality, and that chat logs reflect not only the outcome but also the thought process.