AI in health faces legal and trust checks

A federal court ruling is colliding with a healthcare push by artificial intelligence companies, raising fresh questions about privacy, accuracy and trust. (money.usnews.com) Reuters reported on April 15 that U.S. lawyers are warning clients not to treat chats with tools like ChatGPT and Claude as confidential after a federal judge in New York refused to shield a former executive’s chatbot exchanges from prosecutors. The case involved Bradley Heppner, former chair of GWG Holdings, who had used Anthropic’s Claude to prepare reports for his lawyers. (money.usnews.com) The ruling sharpened a basic distinction: attorney-client privilege usually protects communications with a lawyer, but not necessarily communications with a chatbot. Reuters said more than a dozen major U.S. law firms have since told clients to use caution and, in some cases, warned that sharing legal advice with a chatbot could waive privilege. (money.usnews.com) At the same time, OpenAI is expanding deeper into medicine with products aimed at both patients and hospitals. On January 7, OpenAI announced ChatGPT Health, and on January 8 it launched OpenAI for Healthcare for organizations that need tools built around Health Insurance Portability and Accountability Act compliance. (openai.com 1) (openai.com 2) OpenAI said ChatGPT Health lets users connect medical records and wellness apps, including Apple Health, Function and MyFitnessPal, so answers can be grounded in their own data. The company said health conversations in that product are kept in a separate space and are not used to train its foundation models. (openai.com) For hospitals and clinics, OpenAI said ChatGPT for Healthcare is already rolling out at AdventHealth, Baylor Scott & White Health, Boston Children’s Hospital, Cedars-Sinai Medical Center, HCA Healthcare, Memorial Sloan Kettering Cancer Center, Stanford Medicine Children’s Health and the University of California, San Francisco. The company said the enterprise product is designed for clinicians, administrators and researchers and includes cited answers drawn from medical literature and guidelines. (openai.com) Fierce Healthcare reported that the enterprise workspace includes templates for discharge summaries, patient instructions, clinical letters and prior authorization support. The publication also noted that OpenAI’s announcement did not say whether guardrails were in place to prevent hallucinated citations. (fiercehealthcare.com) Patients are already using consumer chatbots before the healthcare system has settled those questions. A West Health-Gallup Center survey released April 15 found that 25% of U.S. adults have used an artificial intelligence tool or chatbot for health information or advice, based on responses from more than 5,500 adults surveyed from October through December 2025. (benzinga.com) Among people who used artificial intelligence for health in the prior 30 days, 59% said they used it to research symptoms before seeing a doctor and 56% said they used it after a medical appointment. The same survey found 61% of those users turned to ChatGPT or Microsoft Copilot, while only 4% said they strongly trust the accuracy of artificial intelligence-generated health information. (benzinga.com) The survey also found sharper pressure at the low end of the income scale: 32% of households earning under $24,000 said they used artificial intelligence because they could not afford a doctor visit, versus 2% of households earning $180,000 or more. An estimated 14 million adults said they skipped a provider visit after getting artificial intelligence guidance. (benzinga.com) That leaves healthcare artificial intelligence moving in two directions at once in April 2026: companies are adding products that promise tighter privacy controls and hospital-grade safeguards, while courts and lawyers are warning that not every chatbot conversation carries the protections users may assume. (openai.com) (money.usnews.com)