Legal experts across the United States are issuing a stern warning to their clients: stop treating AI chatbots like a private confessional. Prominent law firms have begun updating their client engagement letters to explicitly forbid discussing sensitive case details with generative AI models like ChatGPT, Claude, or Gemini. The core of the issue lies in the potential destruction of Attorney-Client Privilege . Under U.S. law, communications between a lawyer and a client are protected from disclosure, but this protection generally only applies when the conversation remains confidential. By inputting case facts into a third-party AI platform, clients may be legally considered to have “waived” that privilege, potentially allowing opposing counsel to subpoena those chat logs during discovery.
Beyond the waiver of privilege, lawyers are raising alarms regarding data retention and training policies. Most AI service providers, unless specifically configured under enterprise-grade privacy agreements, store user prompts to further train their models. This means a client’s trade secrets, litigation strategies, or admissions of liability could theoretically resurface in the AI’s future outputs or be accessed by the tech company’s employees during “human-in-the-loop” reviews. Legal analysts point out that while a human assistant is bound by confidentiality, an AI “black box” offers no such ironclad guarantee. As courts begin to grapple with how AI interactions fit into existing evidentiary rules, the prevailing advice from the American Bar Association (ABA) is simple: if you wouldn’t say it in a crowded elevator or post it on social media, don’t type it into an AI prompt.
