Two AI agents outscored emergency doctors on real cases, then ordered twice the bloodwork.

Run time: 5:31

In today's episode:

1. Two AI agents match or beat doctors in Nature
2. Google's AMIE outscores physicians on disease management
3. AI flags small pancreatic tumors on routine CT
4. Joint Commission rolls out first hospital AI governance cert
5. Doctors split on patients reading scans with AI
6. Ambient AI scribes head for every VA hospital
7. Anthropic says Fable and Mythos return within days
8. Researchers reframe AI agents as expanding, not replacing, engineers

TL;DR:

• Two landmark Nature papers (June 17) put autonomous AI agents at or above doctor level — but on constructed cases with known answers, no real patients, and one padded its lead by over-ordering tests.
• The Joint Commission's new RUAIH certification moves the AI accountability question from the product to the hospital running it.
• Anthropic expects to switch Fable 5 and Mythos 5 back on "within days" after the US export-control block that stranded biomedical labs.

Sources cited:

• Nature
• Science Media Centre expert reaction
• Topol's Ground Truths commentary
• Nature
• AuntMinnie / Radiology
• Healthcare IT News
• AuntMinnie
• npj Digital Medicine
• Korea JoongAng Daily
• The Decoder

Subscribe: YouTube

medAI Times is for educational and informational purposes only. The content does not constitute medical advice, diagnosis, treatment recommendation, or professional clinical guidance. Consult qualified healthcare professionals and refer to official sources before making clinical, research, regulatory, or business decisions.