Who's listening, and who's deciding

This letter explains, once a week and in plain language, how artificial intelligence (AI) is showing up in health care and what you can ask about it. It gives general information, not medical advice. For questions about your own care, talk with your doctor.

An AI may be taking notes at your next visit

Some doctors now use AI note-takers. The tool listens to the visit and writes a draft of the visit notes, which the doctor or nurse reviews and corrects before signing (Department of Veterans Affairs).

The Department of Veterans Affairs (VA) uses them widely. It says that by June it had given the tool to all of its primary care providers, including doctors, physician assistants and advanced practice nurses. It uses the tool with a veteran's spoken permission. "Veterans can opt out at any time, even mid-visit," it says (VA). On Sept. 22, Abridge said the VA had chosen it to supply the tool under a new contract. Abridge was one of two companies in the VA's test of the tool. The other, Knowtex, told the news site Nextgov it was chosen too. The contract runs five years. It can pay up to about $776 million in total to all the companies on it (Abridge; Nextgov). Clinics outside the VA are adding them too. In Mississippi, a state award paid for with federal money will give AI note-takers to about 500 doctors and other providers at Hattiesburg Clinic (Mississippi governor's office).

The rules differ from state to state. In Colorado, a therapist must tell you in writing before AI records or transcribes a therapy session. The therapist must also get your written permission, and cannot turn you away if you say no (Colorado Division of Professions and Occupations).

What you can ask: "Is an AI tool listening today?" "Can I say no?" "Will you check what it writes?" Under HIPAA, the federal health privacy law, you have the right to see and get a copy of your health records. You can also ask to have corrections added when something is wrong (Department of Health and Human Services).

Insurers say hospitals' AI billing tools are raising costs

When you stay in a hospital, the conditions written in your record help decide how much the hospital is paid. Many hospitals now use AI to help turn those records into billing codes (Fierce Healthcare).

On Sept. 24, the Blue Cross Blue Shield Association, the national group of Blue Cross and Blue Shield insurers, released a study. It said the share of hospital stays billed as more complicated rose from 37% to 40% between early 2023 and the end of 2025. It found no matching change in the care patients got. It blamed hospitals' AI billing tools and estimated the extra cost to its plans at about $942 million over two years. The association said its study was limited because it relied on billing claims, not medical records. It also said spending like this raises premiums and out-of-pocket costs (Fierce Healthcare).

Hospitals see it differently. The American Hospital Association, which represents hospitals, says patients are older and sicker than before and that record-keeping has improved. Hospital leaders have said AI helps them record care accurately and respond when insurers deny claims (Medical Daily).

What you can do: After a hospital stay, you can ask the hospital's billing office for an itemized bill that lists each charge. If your bill or your record lists a condition you don't recognize, ask the billing office or your doctor what it refers to.

Medicare's AI review pilot draws complaints

In six states, traditional Medicare is testing an extra review for some procedures. The doctor or hospital can ask for approval in advance, which is called prior authorization, or have the claim reviewed after the procedure and before Medicare pays. The procedures include certain nerve stimulator implants, skin and tissue substitutes, and knee arthroscopy (surgery through small cuts, with a camera) for osteoarthritis of the knee. The pilot, called WISeR (short for Wasteful and Inappropriate Service Reduction), uses companies that review requests with help from AI. It covers Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington, and runs through 2031. It does not apply to Medicare Advantage, the private plans many people choose instead of traditional Medicare. Medicare says every recommendation not to pay comes from a licensed clinician, and that "Health care coverage for people with Medicare does not change" (Medicare).

The Electronic Frontier Foundation, a digital rights group, sued to get records about the pilot. The records show that two of the companies denied 5,944 requests in the first three months. One company, Virtix, denied more requests than it approved in that period, the foundation said. One request went unanswered for 83 days (Electronic Frontier Foundation). On Sept. 24, the Medicare Rights Center, a patient advocacy group, said the pilot "should be terminated immediately" (Medicare Rights Center).

What you can do: If you have traditional Medicare in one of these states and are waiting for approval of a procedure, ask your doctor's office where the request stands. You can also call Medicare at 1-800-MEDICARE (1-800-633-4227).

What privacy law covers when you use a chatbot

If you ask an AI chatbot about symptoms, test results or medicines, it helps to know which privacy law applies. HIPAA covers your doctors, hospitals, health plans and the companies that work for them (HHS). It generally does not cover an app or chatbot you choose to use on your own. The Department of Health and Human Services puts it this way: "In most cases, unless the app is provided to you by a covered entity or its business associate, the HIPAA Rules also do not protect the privacy of data you've downloaded or entered into mobile apps for your personal use" (HHS). So what happens to what you type depends mostly on the company's own policies. Some state laws and federal rules add protections. For example, the Federal Trade Commission requires some health apps to tell users about data breaches (FTC).

California may change that for its residents. A bill on Gov. Gavin Newsom's desk, Assembly Bill 1979, would make companies that offer health care chatbots to consumers follow the state's medical privacy law. He had not acted on it as of Sept. 28 (CalMatters Digital Democracy), and he has until Sept. 30 to sign or veto it (National Nurses United).

Health plans are using AI to reach patients, too. Cigna said on Sept. 23 that it will use OpenAI's AI models to help its cancer nurses and case managers. The tools will bring together a member's medical, pharmacy, behavioral health and benefits information. Cigna says patients' own questions and reported side effects can also help show what support helps most. The company says its clinicians "remain responsible for patient care and clinical decisions," and that "no patient or member data is maintained by OpenAI or used by OpenAI for model training" (Cigna; Fierce Healthcare).

What you can ask: Before you type health details into a chatbot, look for its privacy policy. Does the company keep what you type? Does it use it to train its AI? And if a nurse from your health plan calls, you can ask why they are calling and what information they used.