Consumer health AI raises privacy and trust issues

A lot of Americans are now asking a chatbot about a rash, a lab result, or a panic attack before they ask a doctor, and a March 25, 2026 Kaiser Family Foundation poll found 32% of adults used artificial intelligence chatbots for health information in the past year. The same poll found 16% asked about mental health, and 58% of those users did not follow up with a clinician afterward. (kff.org) People are not doing this because they think the bots are perfect. A Pew Research Center report published April 7, 2026 found 22% of Americans get health information from artificial intelligence chatbots at least sometimes, and most users rated them as more convenient than accurate. (pewresearch.org) The appeal is simple: a chatbot answers in seconds, in plain English, and never puts you on hold. Kaiser Family Foundation found 65% of users said speed was a major reason, 41% used artificial intelligence before deciding whether to see a provider, and 36% said they felt more comfortable looking up health information privately. (kff.org) The pressure behind that behavior is the U.S. care system itself. Mashable reported on April 9, 2026 that Americans owe more than $220 billion in medical debt, that around half of Americans struggled to afford health care last year, and that the country is projected to need 114,000 more physicians by 2028. (mashable.com) That is why companies are pushing beyond old symptom checkers into tools that read your records and talk back like a guide. Medical Economics reported on February 11, 2026 that patient tools such as ChatGPT Health, Amazon One Medical’s Health AI, and Anthropic’s Claude for Healthcare are moving closer to lab results, patient portals, and inboxes. (medicaleconomics.com) The change is not just that the software answers a question. Medical Economics said newer tools act more like a narrator of health data, turning a cryptic lab value into a sentence that sounds authoritative enough to shape what a patient thinks before the visit even starts. (medicaleconomics.com) That authority is where the privacy fight starts, because the most useful answers often require the most personal inputs. Mashable reported that OpenAI launched ChatGPT Health in January 2026 as a free consumer tool that encourages users to upload medical histories for tailored guidance, while privacy advocates warned it expands a market for sensitive data outside the usual medical setting. (mashable.com) A lot of users assume health data is protected the moment it looks medical, but United States law does not work that way. The Federal Trade Commission says the Health Insurance Portability and Accountability Act usually does not apply to consumer health information held in an app that is not offered by a covered health provider or its business associate, even if the data originally came from a doctor or hospital. (ftc.gov) Federal regulators have already been trying to close that gap. The Federal Trade Commission finalized changes to its Health Breach Notification Rule on April 26, 2024, with the updated rule taking effect on July 29, 2024, and the agency said the changes were meant to clarify that health apps and similar technologies not covered by the Health Insurance Portability and Accountability Act can still trigger breach-notification duties. (ftc.gov, federalregister.gov) That still leaves a messy middle where a patient may paste test results into a chatbot, get a confident summary, and never know which rules apply to storage, sharing, or deletion. The Department of Health and Human Services says that once health information is sent to an app at the individual’s direction, and that app is not a covered entity or business associate, the information is no longer protected by the Health Insurance Portability and Accountability Act rules. (hhs.gov) Clinics now have a second problem after privacy: patients are arriving with machine-written interpretations already in their heads. Medical Economics said primary care physicians are already seeing changes in how patients prepare for visits, interpret test results, and form expectations, which means the exam room is becoming partly a fact-checking session for what the chatbot said at home. (medicaleconomics.com) The legal risk is moving in the same direction as the clinical risk. Mashable reported that the Tech Justice Law Project is representing people suing OpenAI over mental health concerns tied to ChatGPT, and privacy advocates told the outlet that products like ChatGPT Health could set up another round of fights over what companies can collect, what they must disclose, and who is responsible when a health chatbot sounds like a doctor but is sold like an app. (mashable.com)