Agents choosing and executing interventions in live animals, with human sign-off on every irreversible step plus a physical kill switch, is one reference design for human-AI collaboration in medicine.
AI NEWS
Health AI News
Chinese and English are the same day. This page lists every item, each with the original link, the author, and the engagement.
Latest day
November 2, 2026
4 items for 2026-11-02, ranked by combined engagement.
Top 3
Unconfirmed Patients accept AI that assists while doctors decide; spelling out human review earns more trust than touting accuracy.
Unconfirmed An early small-sample result worth tracking but not extrapolating; labs and emergency care are likely early landing spots for AI imaging.
All
Exin had Claude, GPT and Grok each perform brain surgery on epileptic mice — all three chose different therapies, mechanisms and brain regionsUnconfirmed
Agents choosing and executing interventions in live animals, with human sign-off on every irreversible step plus a physical kill switch, is one reference design for human-AI collaboration in medicine.
CollapseExin had Claude, GPT and Grok each perform brain surgery on epileptic mice — all three chose different therapies, mechanisms and brain regions
Gabriel Ocaña-Santero, co-founder of biotech startup Exin Therapeutics, says ChatGPT and Grok have now done their first mouse brain surgeries after Claude: each got the same spreadsheet of 19 gene-therapy vectors, the same epilepsy mouse model and a motorized stereotaxic rig with the goal of reducing seizures, and each picked different therapies, mechanisms and brain regions. Tom's Hardware reports the first run used Anthropic's Sonnet: the agent chose the drug and region, derived coordinates from three papers and drove the arms to inject 250 nanolitres, while a licensed vet approved every irreversible step, held a kill switch and performed the craniotomy; Opus 5.5 refused once it detected a biological experiment. The mouse recovered well, but efficacy data has not been released and ethics and accountability concerns are significant.
Pew: 60% of US adults are uncomfortable with doctors using AI for diagnosis and treatment, while 78% are open to AI translation help
Patients accept AI that assists while doctors decide; spelling out human review earns more trust than touting accuracy.
CollapsePew: 60% of US adults are uncomfortable with doctors using AI for diagnosis and treatment, while 78% are open to AI translation help
HealthAINav breaks down a Pew Research Center survey released Oct 29: 3,488 US adults polled June 22–28 (margin of error ±1.8 points). 60% would be uncomfortable with their doctor using AI to diagnose disease or recommend treatment, 30% would be comfortable and 10% are unsure. Per the post, 78% are open to AI helping with translation, while just 3% are comfortable with AI making diagnoses on its own; acceptance rises markedly when human review is included. Pew also found about half think AI would worsen the patient-doctor relationship and health-data security.
Biohub's Remo-ID: 10 µL of blood and ~12 minutes of stain-free imaging sorted 43 of 47 people into sepsis, non-infectious ICU illness or healthy (early preprint)Unconfirmed
An early small-sample result worth tracking but not extrapolating; labs and emergency care are likely early landing spots for AI imaging.
CollapseBiohub's Remo-ID: 10 µL of blood and ~12 minutes of stain-free imaging sorted 43 of 47 people into sepsis, non-infectious ICU illness or healthy (early preprint)
X user devcayer describes Biohub's Remo-ID model: 10 µL of blood, about 12 minutes of imaging and no stains. By his account, the model sorted 43 of 47 people into sepsis, non-infectious ICU illness or healthy without being told which cell features to look for. It is an early preprint with only 47 people, not yet peer-reviewed or externally validated. Sepsis has a tight early-detection window, so a fast stain-free blood test would be clinically valuable if confirmed.
AHA Scientific Sessions holds an AI-in-cardiology late-breaking session on Nov 9: four studies including smartwatch ECG for LV dysfunction and a randomized AI echo trial
Randomized and external validation are becoming the entry bar for medical AI; checkup and cardiology providers should watch the Nov 9 results.
CollapseAHA Scientific Sessions holds an AI-in-cardiology late-breaking session on Nov 9: four studies including smartwatch ECG for LV dysfunction and a randomized AI echo trial
HealthAINav notes the American Heart Association's Scientific Sessions will hold a late-breaking AI-in-cardiology session on Nov 9 with four presentations. They are WATCH-LV (AI smartwatch ECGs to identify left ventricular dysfunction), PROVAR+ (external validation of an AI-ECG approach to structural heart disease), a blinded randomized trial of AI-assisted echocardiogram interpretation, and AI coronary plaque analysis and risk assessment in more than 269,000 patients. Results have not been presented yet, and the author cautions that not all of them will show whether AI improves patient outcomes.