The FDA announced on April 28, 2026 a pilot with AI firm Paradigm Health covering trials run by Amgen and AstraZeneca, with the agency's chief AI officer estimating real-time data flows could shave 20 to 40 percent off trial timelines.
A Cleveland Clinic study of 4,000 clinicians found 60% said an ambient AI scribe increased their likelihood of staying in practice, with 97% reporting overall satisfaction with the tool.
A University of Toronto-led study finds that of 1,357 FDA-cleared AI medical devices, only three have ever been tested on whether they improve patient outcomes like survival or hospitalization.
a2z Radiology AI's newly FDA-cleared tool flags seven acute conditions from a single abdomen-pelvis CT scan, and the startup says it cut radiologist reporting time by 17.8% in testing.
Leica Biosystems' newly FDA-cleared Aperio iQC DX caught up to 24% more slide defects than histotechnicians while cutting manual quality-review time by 69%, the company says.
A Nature paper from Nobel laureate David Baker's lab shows generative AI can design antibodies from scratch with atomic-level precision, bypassing the immune system entirely.
A retrospective Swedish study published in Radiology found that three commercial AI-CAD systems could have flagged subtle breast-cancer signs on mammograms years before radiologists caught them clinically.
The American Nurses Foundation and Hippocratic AI handed three nurse-led teams $10,000 microgrants to study AI's effect on patient safety and documentation — a grassroots counterweight to boardroom AI decisions.
Roche and Recursion Pharmaceuticals advanced the first validated target from their AI partnership on August 5, 2026 — a neurodegenerative disease target mined from a microglia data map of over a trillion cultured neurons, the first of up to 40 programs under their up-to-$12-billion deal.
A Nature Medicine study from Oxford, UCL, and the UK AI Security Institute ran 810 simulated conversations through nine AI chatbots and found mental-health risk builds gradually through 'Vulnerability-Amplifying Interaction Loops' rather than single bad replies.