Judy Faulkner has spent the past year putting AI at the center of Epic’s EHR. At the company’s August 2025 Users Group Meeting, the founder and CEO of the Verona, Wis.-based EHR vendor introduced “healthcare intelligence,” her preferred name for the technology, and said Epic had 160 to 200 AI projects in the works. This August, she unveiled Ergo, the company’s healthcare intelligence platform, along with Agent Factory, an AI agent-building platform offering 120 out-of-the-box AI features.
But three times this year, Ms. Faulkner has pumped the brakes: on blind trust in AI, on predictions that it will replace physicians and, most recently, on Epic’s own product development.
Here is what she said, as reported by Becker’s:
1. AI can be ‘gamed.’
In an April 24 episode of “Freakonomics Radio,” Ms. Faulkner cautioned that AI can be manipulated and that the industry needs to watch it closely.
“Lots of things have been gamed lately. Capitalism has been gamed. Government has been gamed. I think AI we have to watch to make sure it doesn’t get gamed,” she said. “If you want AI to say a certain something, you put it in repeatedly all over the place. It goes back to if you repeat something often enough, people will think it’s true. Well, AI will think it’s true.”
She paired the warning with a case for the technology on the same podcast, citing a customer that told Epic its AI sepsis detection tool cut its cases to zero and saved more than 100 lives.
2. AI won’t replace physicians.
During a June 23 interview at Aspen Ideas: Health in Colorado, Jon LaPook, MD, chief medical correspondent for CBS News, pointed to research suggesting AI can already outperform physicians at diagnosis, and that adding a physician to the mix can degrade the models’ results. Ms. Faulkner pushed back.
“What we hear people say is that the physician plus the AI is better than either of them alone,” she said.
She described AI’s role as surfacing patterns clinicians can miss, pointing to sepsis, and urged health systems to adopt some form of sepsis-detection tool regardless of which vendor supplies it.
3. She halted most of Epic’s development to fix security flaws.
At its August Users Group Meeting, Epic disclosed that it is a participant in Anthropic’s Project Glasswing and a heavy user of Mythos, a restricted-access AI model with advanced cybersecurity capabilities. Epic said the model surfaced vulnerabilities in its several-hundred-million-line codebase that its own developers had missed, and told customers to expect more urgent security patches.
About a month later, Ms. Faulkner said at an industry conference that Epic had paused most new product development to close the security gaps, a plan expected to take about six weeks. Testing with Mythos found that certain software configurations could let someone view sensitive patient records without the access appearing in a digital audit trail, Epic Chief Security Officer Stirling Martin told The New York Times.
Epic later said its new products remain on track.
“Our development roadmap hasn’t changed since we presented it at our August 2026 Users Group Meeting,” an Epic spokesperson told Becker’s.
Ms. Faulkner also indicated that Epic’s development process could look different once the current security work is complete, with cybersecurity becoming a more constant consideration as new threats emerge.
The post 3 times Judy Faulkner pumped the brakes on AI appeared first on Becker's Hospital Review | Healthcare News & Analysis.