Healthcare AI News 1/7/26

News

OpenAI announces ChatGPT Health, which supports health conversations with encryption, isolation from model training, and connectivity to wellness and medical records applications such as Apple Health and MyFitnessPal.

OpenAI reports that 40 million people use ChatGPT each day for health information. Users ask it to help them decipher medical bills, spot charging errors, file insurance appeals, and in some cases diagnose conditions or manage their care. Seventy percent of those conversations take place outside normal clinic hours. The company also cites reports that nearly half of US nurses use AI weekly.

A Wall Street Journal analysis finds that 27% of US health systems are paying for AI tool licenses, triple the average across industries. The strongest results come from labor heavy administrative work such as note taking, triaging patient calls, and processing insurance claims. A health system found that its use of an Epic tool to manage denials reduced the labor required by 23% and increased the percentage of overturned denials. However, the report adds that Mount Sinai halted its use of Epic’s draft reply tool for patient messages after physicians said that its output required excessive rewriting and sometimes contained questionable information.

Utah launches a pilot to allow AI technology from Doctronic to autonomously manage prescription refills for 190 common medications. The company also markets a free chatbot that assesses systems, offers guidance, and then offers to connect users with virtual providers for a $39 virtual consultation. The goal is to reduce primary care wait times, but medical groups warn that physician oversight is needed. The FDA has not reviewed the automation process and could impose regulations on its use.

The FDA will step back from regulating low-risk wellness technologies, including fitness apps and activity trackers. Products that only share information won’t need clearance as long as companies avoid making medical claims.

Business

CVS Health highlights technology, including AI, as central to its strategy. It has rolled out an AI-first consumer engagement platform across CVS Pharmacy, Caremark, Aetna, and its care delivery units to support prescription, benefit, and care navigation. The company suggests it may eventually offer parts of the platform to outside customers.

Research

Researchers develop SleepFM, an AI tool that can predict 130 disease categories using only data from polymonography (overnight sleep studies), including dementia, heart attack, heart failure, stroke, chronic kidney disease, atrial fibrillation, and all-cause mortality. The tool’s accuracy was measured by linking each patient’s sleep record to their EHR data to find occurrence of related events such as coded diagnosis, procedure and encounter histories, mortality data, and the timestamps of clinical events.

Other

University of Colorado Anschutz profiles AIDA, a self-developed AI assistant that summarizes a patient’s Epic chart for radiologists. Aakriti Pandita, MD, assistant professor of medicine and co-developer of the tool, says, “We don’t need AI to help diagnosing patients. We need AI to help the tasks that are repetitive and redundant and administrative in nature.”

A mother whose teenage son died of an overdose says he used ChatGPT to get advice on dosing illicit drugs and how to achieve different highs. The company says he accessed a flawed version of the model that was known to give unsafe health responses and that he sidestepped safeguards by framing his questions as hypothetical. ChatGPT even suggested a music playlist as part of its recommendation that he drink two bottles of cough syrup.

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Follow on X, Bluesky, and LinkedIn.
Sponsorship information.
Contact us.