It has become increasingly common for people to confide in or share sensitive personal information with AI chatbots like ChatGPT. Some use it to unpack relationships, while others discuss sensitive messages or work-related documents. Many have even turned to such services for medical diagnosis. This intimacy, though, creates an uncomfortable new problem. A private ChatGPT conversation isn’t necessarily a legally protected one.
The Washington Post reviewed public records and local reporting and found chatbot conversations cited in 12 civil and criminal court cases over the past two years. The real number could be considerably higher because investigators and litigants do not have to present every piece of evidence they collect in court.
How ChatGPT conversations become a digital paper trail
A Google search might reveal what somebody wanted to know, while a long ChatGPT conversation can dig much deeper. When dropping a prompt on ChatGPT, there is usually background history behind it and some form of explanation. One legal expert interviewed by the Post described chatbot logs as unusually revealing because the entire back-and-forth can make someone’s thought process dangerously explicit.
In one Missouri criminal case cited by the Post, police were allowed by a suspect to search his phone after 17 cars were damaged. They discovered that, around the time of the incident, he had asked ChatGPT whether investigators could determine that he was responsible. The statements were later cited in the case, and he ultimately pleaded guilty to felony property damage.
Civil cases create another route. During discovery, devices and digital records can sometimes be searched for relevant evidence. In one employment dispute, ChatGPT conversations about whether deleted emails could be recovered became part of an argument that evidence had been withheld. You’re not in danger by simply talking to ChatGPT, though evidence on your phone or computer can be used in a case against you.
Private doesn’t mean privileged
You might assume attorney-client privilege while seeking legal advice, or have similar confidentiality with your doctor, who has your medical history. Unfortunately, none of this applies to AI chatbots. OpenAI CEO Sam Altman has argued that sensitive AI conversations should eventually receive protections similar to communications with doctors and lawyers. However, courts have yet to agree.
In February, a federal judge rejected an executive’s attempt to keep conversations with Claude away from prosecutors under attorney-client privilege, reasoning that Claude is not a lawyer and the conversations were not made at the direction of his attorneys. OpenAI has acknowledged this legal boundary, and its current policy does state that it can be compelled to retain information following lawful legal process. In just the second half of 2025, the company received 75 government requests for content, disclosed data in 62 of them, and disclosed data covering 84 accounts.
Temporary Chat isn’t safe either
OpenAI lets you delete individual conversations, while Temporary Chat keeps conversations out of normal history and automatically deletes them within 30 days. But that’s not exactly the case. OpenAI explicitly says information can be kept longer when required for safety or legal reasons — including when the company receives a lawful subpoena.
So, depending on the circumstance, that information can reach the hands of law enforcement. Merely deleting a conversation won’t actually mean you’ve cleared any kind of trail. Keep in mind that this doesn’t imply that all ChatGPT users are at legal risk. However, you might want to reconsider before starting a conversation with AI that has some very sensitive or highly confidential details.
