Califf warns AI in health care 'overhyped'

On a makeshift stage in a Midtown Manhattan office earlier this week, former Food and Drug Administration Commissioner Robert Califf struck a measured tone about the potential for artificial intelligence in health care. Asked whether the technology was overhyped he said it was. "I hear way too much about the money. I'm not hearing a lot of human values coming through discussions," he said. Adding:

"Almost all of the technology is being applied to optimizing the financial status of healthcare delivery entities or companies that are making medical products and that's not aligned with equitable, better patient outcomes. So until someone puts a soul back in the system, I think it's going to get worse and worse."

From a regulatory perspective, Califf cited the safety of the food system, which is guided by rules that are generally followed, as a model. What's needed is a set of guiding principles that allow for "self-regulation" by health systems that are accountable for patient care.

Califf was critical of federal health leadership under Secretary Kennedy and said he was "not optimistic in the next several years that there is going to be any change in the Wild West" for health AI. "I'm expecting there to be some announcements in the near future about more deregulation in this area."

Califf's remarks echo his past statements about the focus on AI-driven profits and who should hold responsibility for AI safety. There is in fact growing use of AI to optimize billing and other administrative tasks which some warn will drive up health care costs. (See below for more on the growth.)

research

How much AI and health IT are hospitals using?

Screenshot 2025-09-18 at 6.37.48 AM As Califf calls on hospitals to conduct rigorous reviews of the tech they use, two new publications provide insight to the scale of hospital adoption for health technologies. The papers published by the health department's IT regulator, ASTP/ONC, and by researchers at the University of Maryland are based on data on health tech use collected annually by the American Hospital Association.

Some takeaways:

The ASTP analysis of 2023 and 2024 AHA data looks at the growing use of artificial intelligence in hospitals. In 2024, 71% of hospitals surveyed reported using predictive AI in their medical record software compared to 66% the year before. More striking is the growth of certain applications (see the chart above) like billing.
Most hospitals (80%) use tech provided by their EHR vendor, while only 50% used third-party tech. Ninety percent of hospitals using the "market leading" EHR (it has to be Epic, right?) use predictive AI. I was surprised that 66% of hospitals have a special committee for evaluating health AI.

Screenshot 2025-09-18 at 6.40.42 AM

Writing in JAMA Health Policy, researchers from UMD mapped AHA data on adoption of health IT like telehealth and health information exchange capabilities to an "area deprivation index" that quantifies the socioeconomic disadvantages experienced by people in different geographic areas. The results show what you might expect: Areas that are better off show more health IT adoption. The ASTP analysis reported a related finding that adoption rates for "AI were lower among hospitals that are small, rural, independent, government-owned, critical access."
The UMD authors say policy remedies may help. For example, their analysis suggests that IT adoption is higher, even in disadvantaged areas, for hospitals participating in accountable care organizations that are incentivized to reduce costs.

regulation

FDA warnings galore

FDA this week published about 100 warning letters to drugmakers and care providers for advertisements or other marketing that regulators deemed misleading. Among those warned are telehealth companies like Hims and Hers. Read more from Lizzy Lawrence and Ed Silverman here.
FDA also issued two warnings to consumers not to use unauthorized devices for monitoring blood pressure and for monitoring vital signs of infants. The former follows an ongoing dispute between the agency and wearable maker Whoop over its blood pressure insights feature that is not cleared by the agency. Of the latter, I never bought such a device for my kids, but the thought that people are using substandard devices to quell anxiety about their sleeping babies.... yikes.

Health tech news roundup

Semler Scientific, maker of a peripheral artery disease test that was widely used by Medicare Advantage will pay $30 million to settle health care fraud claims with the Department of Justice. Read more from Lizzy here.

Surgeon General nominee Casey Means has agreed to divest her holdings in health tech company Levels Health, which she co-founded. Read more about her financial disclosures from Sarah Todd and Isabella Cueto here.
A new report from Peterson Health Technology Institute concludes that virtual treatment offerings for opioid addiction are as effective as usual care and may help increase retention in treatment. The authors are unconvinced the solutions reduce health care costs. Surprisingly, PHTI argues the virtual tools do not increase the number of people getting treatment.
Nutrition startup Fay is t he latest company to join Amazon's Health Benefit Connector that helps users find digital health products that are covered by their insurance. I'm dying to know if anybody is actually getting connected to care through the Amazon program. Fay earlier this year announced it raised $50 million.
Health IT unicorn Innovaccer continued its shopping spree with the acquisition of Story Health to boost its support for specialty care. Story was founded by Verily co-founder Thomas Stanis and and had raised $26 million from investors.

More around STAT

Check out more exclusive coverage with a STAT+ subscription

Read premium in-depth biotech, pharma, policy, and life science coverage and analysis with all of our STAT+ articles.

What we're reading

Parents testify on the impact of AI chatbots: 'Our children are not experiments', CNBC
Are Republicans changing their minds about AI safety?, Platformer
Four reasons why generative AI chatbots could lead to psychosis in vulnerable people, STAT