Physician AI Adoption Surges as mRNA Cancer Vaccine Trial Is Halted

Today's stories trace two intersecting threads in medicine: the accelerating — and sometimes faltering — integration of artificial intelligence into clinical care, and a significant setback for a personalised mRNA cancer vaccine that underscores how high the stakes remain even for promising technology.

Sunday, 30 August: BioNTech and Genentech Halt Phase 2 mRNA Cancer Vaccine Trial

A personalized mRNA cancer vaccine developed by BioNTech and Genentech has been halted after an independent safety board identified a troubling signal in overall survival. The drug, autogene cevumeran (BNT122), was being evaluated as a monotherapy in patients with resected stage 2 and 3 colorectal cancer — people whose tumour had been surgically removed but who remained at high risk of recurrence.

The Data Safety Monitoring Board found a numerical imbalance: more patients in the vaccine arm of the trial had died than in the control group. This kind of asymmetry does not automatically indicate that the vaccine caused harm — Phase 2 trials often carry statistical noise — but standard ethical protocols require an immediate halt pending a full data review when any unexpected survival difference appears. BioNTech shares fell roughly 7.5% on the announcement.

A separate Phase 2 trial combining autogene cevumeran with immunotherapy continues and was not affected. The halt lands in the same week that Moderna reported encouraging Phase 3 data for its own personalised mRNA cancer vaccine in melanoma — a reminder that mRNA oncology is a broad platform encompassing multiple cancer types with very different tumour biology. The two trials should not be read as contradicting each other; they are parallel experiments in distinct disease settings.

Stage of evidence: Phase 2 trial; halted on a numerical survival signal pending data review.

Sunday, 30 August: Two in Three U.S. Physicians Now Use AI Weekly, Doximity Finds

Doximity's 2026 Physician Compensation Report — drawing on responses from more than 23,000 doctors, the largest such survey of its kind — found that over 65% of U.S. physicians now use AI clinically or administratively on a daily or weekly basis. The scale of adoption would have been difficult to predict just two years ago; it now describes the majority of the American medical workforce.

The economic dimension of the findings is equally striking. More than 40% of physicians believe they — rather than hospital systems or insurers — should capture most of the cost savings AI generates in their practices. In 39% of specialties, AI proficiency is already a factor in hiring decisions and performance reviews. Nearly one in four physicians anticipates an AI-linked pay increase within the year.

These findings signal that AI is transitioning from a clinical novelty to professional infrastructure — and that a contested negotiation has begun over who benefits economically from efficiency gains. For patients, the relevant question remains whether faster, AI-assisted workflows translate to better outcomes or lower costs. That evidence has yet to accumulate at scale.

Sunday, 30 August: AMA Survey — Physician AI Adoption Has More Than Doubled Since 2023

The American Medical Association's 2026 Physician AI Survey, conducted with 1,692 doctors across specialties, found that 81% now use AI professionally — more than double the 38% recorded in 2023, the sharpest adoption spike for any clinical technology in recent memory. The average physician today uses 2.3 distinct AI applications, compared with 1.1 three years ago.

The leading uses are administrative: summarising research literature (39% of respondents) and drafting discharge instructions (30%). These are time-consuming clerical tasks rather than diagnostic decisions, suggesting that AI is first taking root where the workflow burden is highest and the risk of error is most manageable.

Confidence appears to be tracking adoption. Seventy-seven percent of respondents said AI improves their capacity to care for patients, up from 65% in 2023. The 12-point gain indicates that early clinician scepticism — typical when any technology is new — is giving way to familiarity for most practitioners. These are self-reported perceptions, not measured patient outcomes, and should be read accordingly.

Sunday, 30 August: First AI-Designed Drug Enters Phase 3 — Rentosertib for Pulmonary Fibrosis

Insilico Medicine's rentosertib has advanced into Phase 3 clinical trials for idiopathic pulmonary fibrosis (IPF), a progressive lung-scarring disease that carries a median survival of three to five years after diagnosis and for which treatment options remain limited. No AI-designed drug has previously reached a Phase 3 pivotal trial.

Rentosertib's significance lies not only in its target — the TNIK kinase pathway, involved in fibrotic tissue remodelling — but in how it was discovered. Both the identification of TNIK as a relevant molecular target and the design of the drug's chemical structure were performed by Insilico Medicine's AI platform, without conventional human-led medicinal chemistry. The molecule is the first end-to-end AI-originated compound to reach the stage at which regulatory approval becomes a realistic prospect if trial results are positive.

For patients with IPF, who currently have access to two approved antifibrotic agents (pirfenidone and nintedanib) that slow disease progression without reversing damage, a new mechanistic class would be meaningful. Phase 3 results are several years away and approval is far from assured — but the milestone reframes the central question in AI drug discovery from "can AI find a candidate?" to "does the drug actually work in patients?"

Stage of evidence: Phase 3 pivotal trial; results expected in several years.

Sunday, 30 August: AMA Launches Four-Module Ethics Curriculum for AI in Clinical Practice

The American Medical Association has introduced a structured continuing medical education (CME) programme on ethical AI use, developed with the Duke Institute for Health Innovation and the Health AI Partnership under the AMA's Code of Medical Ethics. The curriculum arrives as the AMA's own survey data confirm that 81% of physicians now use AI professionally — making structured ethical guidance urgently practical rather than merely theoretical.

The four modules address distinct challenges: auditing AI systems for demographic and diagnostic bias; evaluating the explainability of AI-driven clinical suggestions; protecting patient privacy in the context of ambient scribing tools that passively transcribe consultations; and assigning accountability when AI contributes to a clinical error.

These are not abstract questions. As AI participates more directly in documentation, triage, and care coordination, the profession must decide how to disclose AI involvement to patients, what standard of transparency is owed, and who bears responsibility when AI-assisted decisions go wrong. The curriculum cannot resolve those debates, but it gives clinicians a shared ethical vocabulary and a formal framework within which to navigate them — a necessary foundation as the technology continues to spread.

This article is journalistic reporting for informational purposes only and does not constitute medical advice. Readers should consult a qualified healthcare professional before making any decisions about their health, treatment, or medication.