Private conversations with ChatGPT are increasingly being presented in legal proceedings. It’s important to exercise caution.

      It has become more common for individuals to confide in or share sensitive personal information with AI chatbots like ChatGPT. Some users explore relationship issues, while others discuss delicate messages or job-related documents. Many have even sought such services for medical assessments. However, this closeness leads to a concerning new issue: a private conversation with ChatGPT does not automatically guarantee legal protection.

      The Washington Post examined public records and local news coverage, uncovering that chatbot dialogues were referenced in 12 civil and criminal court cases over the last two years. The actual figure may be significantly higher, as investigators and litigants aren't required to present every piece of evidence gathered in court.

      How ChatGPT conversations create a digital footprint

      A Google search might show someone’s inquiries, but an extended ChatGPT exchange can delve much deeper. When posing a prompt, there's typically background context and some form of elaboration. One legal expert interviewed by the Post noted that chatbot logs are particularly revealing, as the complete back-and-forth can expose an individual’s mindset in a potentially risky manner.

      In one criminal case in Missouri highlighted by the Post, a suspect permitted police to search his phone after 17 vehicles were vandalized. They found that, shortly before the incident, he had asked ChatGPT if investigators could ascertain his culpability. These statements were later referenced in the case, leading to his guilty plea for felony property damage.

      Civil cases present another avenue. During discovery, devices and digital records may be scrutinized for pertinent evidence. In an employment disagreement, conversations with ChatGPT regarding the recovery of deleted emails became part of a claim that evidence had been concealed. While simply chatting with ChatGPT doesn’t put you in jeopardy, any data on your phone or computer can potentially be used against you in legal matters.

      Private doesn’t mean protected

      You might believe that seeking legal advice guarantees attorney-client privilege, or that your doctor-patient confidentiality secures your medical history. Unfortunately, none of this protection extends to AI chatbots. OpenAI's CEO, Sam Altman, has argued for establishing protections for sensitive conversations with AI comparable to those for communications with medical and legal professionals. However, courts have yet to come to a consensus.

      In February, a federal judge denied an executive's effort to shield discussions with Claude from prosecutors under attorney-client privilege, reasoning that Claude is not a lawyer and the exchanges weren't conducted under his attorneys' direction. OpenAI has acknowledged this legal limitation, stating that it may be required to retain information when compelled by a lawful legal process. In the latter half of 2025 alone, the company received 75 government requests for content, complying with 62 of them and disclosing data from 84 accounts.

      Temporary Chat isn’t safe either

      OpenAI offers the option to delete specific conversations, while Temporary Chat keeps discussions out of typical history and automatically erases them within 30 days. However, this isn't entirely accurate. OpenAI explicitly states that information may be retained longer if necessary for safety or legal purposes — including instances when the company receives a valid subpoena.

      Thus, depending on the situation, that information could end up with law enforcement. Simply deleting a conversation doesn’t truly erase any trace. It’s important to keep in mind that not all ChatGPT users face legal risks. However, you might want to think twice before initiating a chat with AI about very sensitive or confidential matters.