Q: does AI waive attorney-client privilege in light of United States v. Heppner? A: In United States v. Heppner, the court held that documents created using the AI tool Claude were not protected by attorney-client privilege or the work product doctrine because Claude is not a lawyer and the defendant used it without counsel’s direction. The opinion also relied on Anthropic’s privacy policy to find confidentiality waived, a step the article criticizes as unnecessary and problematic. Q: What were Judge Jed Rakoff’s core holdings in Heppner about AI and privilege? A: Judge Rakoff held that Claude is not a lawyer and therefore direct communications with it do not establish an attorney-client relationship, and he concluded the work product doctrine did not apply because counsel did not direct the defendant’s use of the AI. He also found that the AI provider’s privacy policy undercut any reasonable expectation of confidentiality, a point the article says goes beyond what was necessary to resolve the case. Q: Why is relying on an AI provider’s privacy policy to determine waiver controversial? A: It is controversial because courts typically assess a user’s reasonable expectation of confidentiality rather than treating vendor terms of service as dispositive. The article warns that elevating privacy policies risks sweeping ordinary cloud-based legal tools into waiver and conflicts with frameworks like ABA Formal Opinions 477R and 512. Q: Can the work product doctrine protect materials created with AI tools? A: The work product doctrine can protect materials prepared in anticipation of litigation or at counsel’s direction, but in Heppner the defendant’s undisputed, self‑directed use of Claude meant the doctrine did not apply. The article also notes that a different magistrate judge in the Eastern District of Michigan held that a self-represented litigant’s ChatGPT exchanges could be protected as work product. Q: What practical steps can lawyers and clients take to reduce the risk that AI use leads to waiver? A: Counsel should direct client AI tasks tied to a legal matter, document that the work is for legal advice or litigation preparation, and keep a record of those directions to show purpose and control. They should also use enterprise-grade or firm-controlled AI with no-training modes, disable chat history where possible, store privileged material on approved platforms with DLP and audit logs, and perform vendor due diligence consistent with ABA guidance. Q: Should courts treat interactive AI chatbots differently from other cloud-based legal tools for confidentiality purposes? A: The article argues courts should not treat interactive AI chatbots differently from other cloud-based tools and should assess confidentiality based on reasonable steps taken to preserve secrecy, not on the form of the interaction. It criticizes Heppner’s framing that emphasized “communicating with” a chatbot and warns that such distinctions do not reflect the underlying technical realities of how services protect data. Q: How might Heppner affect clients with limited resources who use AI translators or summarizers? A: The decision could disadvantage clients who cannot afford human intermediaries like interpreters, translators, or paid consultants because they may lose privilege for low-cost AI assistance. The article warns this outcome could entrench resource disparities by making access to confidential legal support depend on a client’s ability to pay for privileged human services. Q: What should future courts and bar associations do to clarify how AI use affects privilege? A: Future courts should resolve AI-related privilege disputes on established grounds—recognizing that AI is not a lawyer and that work product requires counsel direction—while determining confidentiality by the facts, purpose, and reasonable steps taken to protect information rather than by vendor terms of service. Bar associations should issue guidance on how attorneys can safely direct clients’ use of AI and advise clients consistent with ABA Formal Opinions 477R and 512.