Why ‘shadow AI’ is becoming a growing issue at health systems

Modern Healthcare – Tuesday, January 20, 2026
By Gabriel Perna

Clinicians are using artificial intelligence on the job, whether the tools are endorsed by their employers or not, creating a challenge for health systems trying to oversee its implementation.

Despite an abundance of vendor solutions, health systems are not adopting and investing in AI solutions fast enough for a clinician workforce eager to use the tools. As a result, some clinicians are using AI in an unauthorized manner, which has sped up the need for more comprehensive governance.

“I can’t think of a technology tool in healthcare that clinicians actually have really wanted to use this badly,” said Dr. Kedar Mate, chief medical officer and co-founder of Qualified Health, which works with health systems to implement AI.

An October survey by Offcall, a company focused on reducing physician burnout, found 67% of 1,000 physicians across more than 200 specialties use AI daily in their practice.

Its usage is only expected to grow among younger physicians.

“Most [medical students] grew up as digital natives. I had a chance to work with some of them at UCLA … every one of them had ChatGPT and OpenEvidence open on their phone. It’s part of their workflow,” said Dave Reese, chief technology officer at Amgen, during a panel at last week’s J.P. Morgan Healthcare Conference.

Health system executives may not realize how much “shadow AI,” a term for AI tools not permitted within an organization, is happening on hospital campuses and within clinics, said Dr. Peter Bonis, chief medical officer at Wolters Kluwer Health, which operates an AI-enabled clinical decision support tool called UpToDate.

“The issue of shadow AI is that there are AI applications that are being used for patient care one way or another by healthcare professionals that are not strictly sanctioned or authorized by that enterprise,” Bonis said. “No one may have been told not to use it, but no one’s sanctioning them.”

Wolters Kluwer Health surveyed more than 500 healthcare professionals, both physicians and administrators, for an upcoming report and found 40% of respondents have encountered an unauthorized AI tool in their organizations, Bonis said.

Some health system leaders are trying to curb the practice. Hackensack Meridian Health, for instance, does not permit certain AI tools, including as ChatGPT, to be used at its facilities.

The Hackensack, New Jersey-based system has a relationship with Google, which has a large language model that is trained for clinical use. Physicians may not like the restrictions but they understand why they’re in place, said Joel Klein, chief digital and transformation officer.

“They understand that bots hallucinate and don’t protect our patients information, they have no place on our organization,” Klein said. “We have data that shows that people actually love the tools that we’ve put into place for them. And they use them pretty aggressively.”

The growing usage of unauthorized AI within health systems requires comprehensive oversight that starts with top leadership and includes employee training, clear communication and ensuring AI is woven into all processes, experts said.

“Governance can’t be an afterthought,” said Dr. Justin Norden, co-founder and CEO of Qualified Health. “It’s not ‘we meet once a month as a committee and that’s AI governance.’ That’s the status quo.”

An analysis published earlier this month by health consultant Manatt Health found the health systems that have received buy-in from clinicians and a return on investment from AI are the ones that have a strategy overseen by their CEO.

Advocate Health, a Charlotte-based health system, has created a comprehensive governance structure that has allowed it to scale AI across the enterprise, said Chief Digital and AI Officer Andy Crowder. The system has used AI to retain staff, reduce clinical burnout and improve outcomes.

Leadership across all levels of the organization is involved with implementing the governance strategy, Crowder said. “Having the right strategy and organizational alignment around AI is really important,” he said.

Access this article at its original source.