Daily Update for September 19, 2026

Show Notes

FDA keeps radiology AI under 510(k) review; telehealth coalition seeks federal licensure path; UW Health sorts 100 new AI features by risk; Angle Health raises $600 million

The voices on this episode are generated, and so is the script, from the day's post at Physician in the Loop. Every fact in the post links to its source.

Read the post and its sources

The Food and Drug Administration (FDA) on Thursday published a final order denying Harrison.ai's petition to exempt experienced manufacturers' new radiology AI from 510(k) review, keeping all four device categories under premarket notification. The agency denied the request in April; the Sept. 17 order in the Federal Register made the denial formal, took effect the same day and applies to every company in the four categories, not only the one that asked. The order followed the FDA's August discussion paper on generative AI, which proposed testing models the way clinicians are tested, and it leaves the premarket gate for radiology AI in place while the agency develops its approach to generative models.

A coalition led by Johns Hopkins Medicine and ATA Action, the American Telemedicine Association's advocacy arm, is seeking federal legislation to create a narrow pathway around state medical licensure, on which every multistate remote care model depends. A Harris Poll of 1,514 direct-care workers found fewer than half comfortable using AI and 39% reporting no effect on their work, a year after the same survey found 89% calling AI skills critical. At a Becker's conference, Yale New Haven Health said it will not launch an AI pilot unless an executive sponsor commits to paying for it if it succeeds, and Rush's chief information officer said vendor pricing has moved to token consumption with no common way to model the cost. An AI-native insurer for small businesses raised $600 million. A fourth California health-AI bill is awaiting the governor's action, and the governor on Friday issued an executive order about kill switches for frontier models that does not mention hospitals.

No new journal evidence on clinical AI met the threshold for inclusion since Thursday. The evidence section returns Monday.

In this episode

• FDA final order keeps all four categories of radiology AI under 510(k) review, denying Harrison.ai's bid for a track-record exemption
• Johns Hopkins, Teladoc, Talkiatry and ATA Action form coalition seeking a federal pathway around state medical licensure
• Harris Poll of 1,514 direct-care workers finds 59% likely to seek a new job, 48% comfortable with AI and 39% seeing no effect from it
• Yale New Haven ties AI pilots to advance funding commitments; Rush CIO says no common way exists to model agent costs across platforms
• UW Health reviews more than 100 AI features that arrived in a single platform upgrade, sorted into low, medium and high risk
• AI-native small-business insurer Angle Health raises $600 million at a $2.7 billion valuation, with "automated clinical steering" in its pitch
• Fourth California health-AI bill on governor's desk, SB 503, would require physician offices to "regularly monitor" decision support AI for bias

What is Physician in the Loop?

How technological innovation is changing medicine for patients and the people who practice on the front lines.

Nora: This is Physician in the Loop, the daily briefing on AI in medicine. I'm Nora Davis.

James: And I'm James Hudson. It's Saturday, September 19th, 2026.

Nora: The FDA just told Harrison.ai no, and it's a no for every company in the category, not just the one that asked. Thursday's final order keeps all four device categories of radiology AI under 510(k) review.

James: Even for companies with a track record already.

Nora: I've got that FDA order, plus Yale New Haven and Rush on what AI pilots actually cost to run.

James: And I've got a coalition pushing for a federal pathway around state medical licensure, plus UW Health sorting a hundred new AI features by risk.

Nora: So the FDA order. Docket FDA-2025-P-5560, published in the Federal Register on September 17th, took effect the same day.

James: Remind me what Harrison.ai wanted.

Nora: A partial exemption from premarket notification, since they already hold nine FDA clearances across 13 indications. The FDA said no back in April, and this order just made it formal.

James: What was the reasoning?

Nora: The April letter said holding a clearance may not reflect that a manufacturer is proficient in what the next device needs. It also said a detection tool and a diagnosis tool aren't the same thing.

James: And the radiologist can't just catch a bad output?

Nora: That's the other piece. The order says the truth about an image may not be routinely available before a faulty output leads to a misdiagnosis.

James: Did Harrison.ai push back?

Nora: They said AI touches about 1% of U.S. diagnostic radiology work here, their own figure, and argued the fix is letting standalone performance studies substitute for multi-reader trials.

James: The FDA didn't bite.

Nora: No. This came right after the FDA's own August paper on testing generative models the way clinicians are tested, so the agency is clearly still working out its approach.

James: And more than 45 comments came in on the docket.

Nora: Most of them against the petition. So the premarket gate for radiology triage and detection tools stays exactly where it was.

James: Alright, my turn. ATA Action, the American Telemedicine Association's advocacy arm, announced a coalition on September 17th with Johns Hopkins Medicine co-chairing.

Nora: Seeking what, exactly?

James: Federal legislation for a narrow pathway around state medical licensure. Follow-up care with an established clinician, specialty and rare-disease care, second opinions, federally regulated trials, and what the release calls technology-enabled services.

Nora: States keep the rest?

James: Standards of care, discipline, malpractice, all stay with the states. Kyle Zebley, who runs ATA Action, said, quote, this is not about federalizing medical licensure.

Nora: No bill drafted yet, though.

James: Not yet. But it matters for AI-first care companies operating across state lines. Right now the number of licensed physicians they need is set by state requirements.

Nora: A federal pathway for technology-enabled services would shrink that number.

James: Exactly. And the Federation of State Medical Boards said back in August that AI isn't ready for a license of its own.

Nora: The coalition isn't asking for one.

James: Right. Just a narrower human path around the boards.

Nora: Okay, back to me. At the Becker's conference, Yale New Haven Health's Lee Schwamm said they're not evaluating a pilot unless the executive sponsor commits to paying for it if it succeeds.

James: That's a real gate before anything even launches.

Nora: It is. He also said Yale New Haven pushes back on vendor deals that pass frontier-model costs through with a markup.

James: And Rush had a different complaint.

Nora: Rush's CIO, Jeff Gautney, said vendor pricing has moved from subscriptions to token consumption, and every platform runs its own console. He said, quote, there's not a common way across those platforms to really do financial modeling.

James: So the clinical question isn't even the first hurdle anymore.

Nora: University of Utah Health's Donna Roach put it simply. Her clinicians say give it to me, quick, quick, quick, and finance says pump the brakes.

James: About 70 use cases in development there, and 20 running, right?

Nora: That's what she said. Meanwhile Rush has spent 18 months building patient-facing agents just for hours, directions and refill requests.

James: So under Schwamm's rule, a documentation tool that hands time back to physicians only survives the pilot if someone already agreed to pay for it.

Nora: That's the standard now, at least at Yale New Haven.

James: Last story's mine. UW Health's Frank Liao told Becker's that one single platform upgrade brought in more than 100 AI features, spanning several software versions.

Nora: A hundred at once. How do you even review that?

James: They sorted them into low, medium and high risk tiers. Only a handful landed in the top tier.

Nora: Who reviews them?

James: Governance widened to include human resources and administrative leaders, because Liao said the features no longer stop at the clinic door.

Nora: Who's on the hook if one of them goes wrong?

James: He said responsibility for outcomes sits with the business unit that uses the tool. He also said governance actually allowed us to go faster.

Nora: UCLA Health's Paul Lukac is doing something similar.

James: Right, a biweekly AI review, about nine months in now.

Nora: So at UW Health, it's not the departments deciding which features get switched on.

James: No. It's the governance review that makes that call, every time a new version rolls in.

Nora: Sources and corrections are at physicianintheloop.org. I'm Nora Davis.

James: I'm James Hudson. See you tomorrow.