Daily Briefing for October 1, 2026

Show Notes

Newsom signs two clinical AI bills, vetoes AI therapy limits; autonomous AI clears skin referrals; ARPA-H funds ICU digital twins; UVA cuts scribe hours

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

Gov. Gavin Newsom signed two bills Sept. 30 that protect licensed clinicians' judgment over clinical AI tools in California and vetoed two others, including one that would have limited AI in psychotherapy.

The signed bills, AB 1979 and SB 503, take effect Jan. 1, 2027, and apply to physician offices and group practices as well as hospitals. The vetoes ended AB 2575, which would have protected nurses and other clinicians who override AI output, and SB 903, which passed the Senate without opposition. In the United Kingdom, an autonomous AI system discharged a quarter or more of urgent skin cancer referrals at two hospitals without clinician review, and missed six cancers, according to a congress presentation.

The Advanced Research Projects Agency for Health funded critical care digital twins at the University of Vermont and Stanford and a Stanford monitor for autonomous cardiology agents. UVA Health cut emergency department scribe hours as its ambient documentation tool spread, the student newspaper reported, and the American Hospital Association asked senators to extend HIPAA-level standards to AI vendors and help rural hospitals hire cybersecurity staff.

In this episode

• Newsom signs AB 1979 and SB 503 on clinical AI, vetoes AB 2575's override protection and SB 903's limits on AI in psychotherapy
• Autonomous AI discharges 25% to 31% of urgent skin cancer referrals at two NHS hospitals, freeing about 8,500 appointments; six cancers missed
• ARPA-H awards up to $79.4 million for ICU immune digital twins and $14.9 million for a monitor of autonomous cardiology agents
• UVA Health cuts emergency department scribe hours after a one-third budget reduction as Microsoft's DAX spreads
• AHA asks Senate to hold AI vendors outside HIPAA to its standards and to fund cybersecurity staff for rural hospitals
• PHTI survey: 68% of employers and 55% of health plans tie digital health payments to performance; 71% of systems use AI documentation
• Family doctors who disclose AI use are rated less warm and less competent in 1,030-person vignette study, most among adults under 45
• Heidi releases agents that pre-chart schedules, draft referrals and sweep panels for overdue screenings, pending clinician approval
• Ortet launches as a health AI lab with a $500 million commitment from Thoreau and NYU's Kyunghyun Cho as CEO

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 Thursday, October 1st, 2026.

Nora: A follow-up on yesterday's watch item: Newsom's deadline came and went, and he acted on all four health AI bills.

James: Signed, vetoed, or a mix?

Nora: A mix. Also today: an autonomous AI system reading skin cancer referrals in the U.K., ARPA-H funding ICU digital twins, and UVA Health cutting scribe hours. Newsom signed AB 1979 and SB 503 on Tuesday, and vetoed AB 2575 and SB 903.

James: Start with what he signed. What do those two actually require?

Nora: AB 1979 makes health facilities, clinics and physician offices take reasonable steps so licensed professionals can still exercise independent judgment when clinical decision support AI informs care. It bars using AI to direct unlicensed staff in, or independently perform, work that requires a license. And it treats companies offering consumer health chatbots as providers under the state's medical confidentiality law.

James: That chatbot piece is new. So a consumer app giving health advice now has the same confidentiality duties as a clinic?

Nora: That's what the bill says. SB 503 is narrower: developers and users of clinical decision support AI have to make reasonable efforts to find and reduce biased impacts, and users have to monitor the tools regularly. Both take effect January 1st, 2027.

James: And the two he vetoed, the override protection and the psychotherapy limits, those just didn't make it.

Nora: Right. Sandy Reding, president of the California Nurses Association, put it this way: “The governor knows A.I. should never replace a nurse's clinical judgment, but he vetoed the bill that would have protected us for using our judgment.”

James: So he signs a bill protecting clinician judgment, then vetoes the one protecting clinicians who actually exercise it.

Nora: That's the tension her statement points to.

James: In the U.K., an AI system reviewed more than 8,300 urgent skin cancer referrals at two hospitals, without a clinician looking at them first.

Nora: And it just discharged people, no second look?

James: At one hospital it discharged 31% of referrals, at the other 25%, over 16 months. Dr. Lucy Thomas presented it at the European dermatology congress. The team says it freed about 8,500 appointment slots and nearly 2,900 clinician hours, and cut the biopsy rate to 27% from 43%.

Nora: What about accuracy? That's the number that matters here.

James: Sensitivity was more than 98% for invasive melanoma and the two common skin cancers, specificity about 72%. But surveillance afterward found six missed cancers, five basal cell carcinomas and one melanoma in situ. No adverse outcomes were reported, and 86% of patients had agreed to let the AI decide alone.

Nora: Six missed cancers is still six people. Has this been published anywhere, or just presented?

James: Just a congress presentation so far, no journal paper yet. The device does carry a European CE mark as a Class III device.

Nora: ARPA-H put real money behind ICU digital twins this week.

James: How much are we talking?

Nora: Up to $38 million to the University of Vermont, with Dr. Gary An leading a five-year project to simulate a patient's immune response and test therapies on the simulation first. VTDigger reported that wider clinical use is still seven to 10 years off.

James: And Stanford got something too?

Nora: Up to $41 million for its own immune digital twin, called IMPRINT, to predict ICU outcomes. Separately, Stanford got about $15 million under a different ARPA-H program to build STEWARD, a monitor for autonomous AI agents in cardiology, with OpenAI, the University of Washington and Tuple Health.

James: A monitor for the autonomous agents themselves. That implies nobody's watching them properly yet.

Nora: Dr. Roxana Daneshjou, a co-leader, said the goal is making sure autonomous agents “remain safe, transparent and accountable.” The researchers told Becker's that no system today does regulatory-grade monitoring of autonomous health agents.

James: So the oversight tool is being built at the same time as the thing it's supposed to oversee.

Nora: At UVA Health, emergency department scribe hours got cut after the program's budget dropped by a third.

James: Is that tied to the AI scribe tool they adopted?

Nora: UVA Health brought in Microsoft's DAX Copilot in 2025. The cut was announced to scribes back in April, and a new ED schedule starting September 15th eliminated shifts and cut availability. The Cavalier Daily reported it Wednesday.

James: Does anyone there connect the two directly?

Nora: Dr. Sarah Wendel, an assistant professor of emergency medicine, said, “I think that this AI tool came into place, and then we, as a kind of a healthcare system budget, we always have to make cuts.” She also said she wasn't the one who made the budget decision.

James: Is DAX even replacing the scribes in practice?

Nora: Not really. Only about two of 36 emergency medicine residents actively use DAX when a scribe is present. The paper found no UVA Health statement explaining the reason for the cut.

James: So the hours disappear before the tool's actually doing the job.

Nora: Before we go, something to watch. The Blue Cross Blue Shield Association says hospitals' AI coding tools added $942 million to its plans' costs over two years, and the Medicare Rights Center is calling for ending WISeR.

James: And states are starting to respond to that directly?

Nora: Connecticut now bars AI-only denials and downcoding for 270,000 people on its state plan.

James: One state, one plan population. Worth watching if others follow.

Nora: Second thing: Alabama's HB 272 took effect today. It requires human review of payer utilization decisions and disclosure when AI is used.

James: So on the same day California tightens rules for clinicians using AI, Alabama tightens rules for insurers using it against patients. Two different states, pulling on the same thread from opposite ends.

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

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