‘AI privilege’ remains unclear
Journalism coverage says conversations with health chatbots could pressure courts to consider a new form of legal protection—an 'AI privilege'—but the legal status is still unsettled. Mashable reported that while users treat chatbots like private confidants, the law does not currently give those exchanges the same protections as doctor‑patient communications. That legal ambiguity affects how apps present privacy and confidentiality to users. (mashable.com)
Courts do not currently treat chats with health artificial intelligence tools like protected talks with a doctor or therapist, even as people use them that way. (mashable.com) Mashable reported on April 11 that the gap is raising pressure for some form of “AI privilege,” a legal shield that would block certain chatbot conversations from being forced into court. The article tied that debate to health and mental-health bots, where users often disclose diagnoses, symptoms, trauma, and medication questions. (mashable.com) Federal evidence law does recognize some confidential relationships, including attorney-client and psychotherapist-patient privilege, but it does not list a general privilege for conversations with software. Rule 501 leaves privilege questions to common law and statutes, which means any new protection would have to be built by courts or lawmakers, not assumed by users. (law.cornell.edu) That uncertainty is not hypothetical. In February 2026, Mashable reported on a federal case in which a judge allowed prosecutors to use a defendant’s prompts to a public artificial intelligence tool, finding they were not protected by attorney-client privilege. (mashable.com) Law firms tracking that case, United States v. Heppner, said the court held that communications with a public artificial intelligence model about legal strategy were not covered by attorney-client privilege or the work-product doctrine. Those analyses said the ruling turned in part on the lack of a confidential lawyer relationship with the tool itself. (fenwick.com) Health apps sit in a separate privacy system that can look stronger than it is. The Federal Trade Commission says many health apps are not covered by the Health Insurance Portability and Accountability Act, the federal medical privacy law known as HIPAA, even when they track mood, sleep, fertility, medications, or symptoms. (ftc.gov) Instead, many of those apps fall under consumer-protection rules, including the Federal Trade Commission’s Health Breach Notification Rule for certain personal health record vendors and related companies. In 2024, the commission expanded that rule to more clearly cover modern health apps and unauthorized disclosures of identifiable health data. (ftc.gov; dwt.com) Psychologists have been warning that users may mistake chatbots for licensed care. The American Psychological Association said in a recent health advisory that general-purpose generative artificial intelligence chatbots were not built solely for wellness or mental-health treatment, and its services arm urged federal regulators in 2025 to add safeguards against bots posing as therapists. (apa.org; apaservices.org) Congress is also starting to circle the issue from the consumer side rather than the evidence-law side. Representative Kevin Mullin introduced the CHATBOT Act, House Resolution 7985, on March 19, 2026, to bar artificial intelligence systems from impersonating licensed professionals in medicine, law, and finance. (house.gov) For now, the practical rule is simpler than the legal debate: a chatbot may feel like a confidant, but that does not make it one in court. Until judges or lawmakers create a new privilege, privacy promises in health artificial intelligence products will keep colliding with the limits of existing law. (mashable.com; law.cornell.edu)
