Chatbots Don’t Have Attorney-Client Privilege and Can Be Used Against Defendants
There are ongoing issues surrounding the use of artificial intelligence and the legal community, particularly about traditional safeguards for clients and if they are applicable with the new technology. Generally speaking, if you or a loved one are under investigation or facing criminal charges, one of the most valuable things you have is the ability to speak candidly with your attorney. The attorney-client privilege exists so that you can tell your lawyer the whole truth, including the bad facts, without worrying that the government will one day read your words back to a jury. However, artificial intelligence tools are quietly eroding that protection. Not because anyone intended it, but because the way people use AI does not fit the way privilege law works. As we previously covered, a judge earlier this year ruled that an individuals use of Athropic’s Claude to analyze his legal situation was not protected by attorney-client privilege. In the end, a criminal defendant’s private brainstorming about his own defense became evidence in the government’s hands. Courts are not uniform on this matter yet, with many yet to weigh in on the matter, but regardless no one facing criminal charges should be betting their liberty on which line of cases a judge will follow.
AI Notetakers and Their Inherent Risks
Another problem for privilege emerging is one most people never think about, because the AI can join conversation on its own. AI notetakers, the bots that sit in on Zoom, Teams, and Google Meet calls to transcribe and summarize, turn everything said in a meeting into stored data. That data may be retained indefinitely, used to train models, sold, or could exposed in a breach. In practice, theoretically a notetaker on a call with between a lawyer and their client could disclose that conversation to a third party without the right safeguards in place. That is precisely the disclosure that can defeat privilege. It also creates a permanent, searchable transcript of exactly the discussion you least want preserved, one that captures every offhand remark, every misstatement, and every bad joke, with no context and sometimes with the wrong speaker attached to the wrong quote as the technology is not without its limits.
This technology, along with court’s rulings make clear that anyone facing charges should not type the facts of their case into ChatGPT, Claude, Gemini, Copilot, or any other public AI tool. Not to “understand the charges,” not to check whether your lawyer is right, not to draft a timeline. Assume anything you enter can end up in the government’s hands. Under no circumstances should one let an AI notetaker sit in on any call with your attorney. Check the participant list before every video meeting and object if a bot is present. Do not upload documents such as discovery, police reports, financial records, or any other potentially incriminating or privileged information into an AI tool for summarizing. While AI is certainly useful in many ways, it should be used by your defense team, in a secure environment, at counsel’s direction (if at all), not by a non-lawyer who does not have some of the same protections for privacy. Privilege is not automatic. If you are under investigation or charged with a crime, close the chatbot and call a lawyer who can advise you and keep your information safe from other parties.
