Research

AI beat generalists 83% to 36%. Read before sharing

The fairer test came in at 80% to 47%, and that is the number worth repeating. Plus: NICE puts a clock on five prostate AI tools.

JAMA: clinicians ask AI outside their specialty

270 million questions, half about treatment, often outside the asker's own field. Plus: the MHRA's plan to 2030.

Lancet trial: an AI took the history before the doctor did

57% of clinicians who read the AI's summary said it may have influenced their approach. Plus: why a new algorithm needs a new clearance.

London licenses, Utah prescribes, Riyadh integrates

The government accepts every Commission recommendation on AI regulation, Utah lets software write acne prescriptions, and Abu Dhabi shows a Gulf Arabic scribe.

The scribe that listens also writes the bill

A US insurer traced a billing jump to AI coding tools and named ambient listening, Ireland sets five gates for AI, and the SFDA takes the stage in Riyadh.

The mammogram second reader is now software

A five-year rebuild of NHS screening, a breast-screening trial that ran the arbitration numbers, and the safety body already investigating the last AI rollout.

ChatGPT can now read the patient record

4,363 physician ratings, 27 use cases, and the number nobody converted back into patients

Endoscopists got worse after the AI arrived

The FDA cleared 1,357 AI devices. Three had outcomes

Authorisation is not the same as evidence of benefit.

The MHRA just drew the line on AI scribes

Most of them aren't medical devices. That's not the relief it sounds like.