A radar sensor called Paul has cut resident falls by an average of 66% across 11 Arizona senior communities, but even its Swiss inventor admits he can't fully rule out that better-watched staff, not the AI itself, deserve some of the credit.
Dartmouth psychiatrist Michael Heinz built Therabot, the first generative AI mental health chatbot with published clinical trial results — but a rival psychologist warns that no algorithm can replicate the empathy real therapy requires.
A 23,132-patient study across 11 RWJBarnabas Health hospitals found an AI early warning system tied to automated rapid-response alerts cut in-hospital deaths among high-risk patients by 18%, researchers reported in NEJM AI.
Mayo Clinic's REDMOD AI model found signs of pancreatic cancer on CT scans radiologists had already called normal, detecting 73% of cases a median of 16 months before diagnosis in a nearly 2,000-scan validation study.
Boston University researchers published an antibody-specific AI framework on August 13, 2026, that improved binding-affinity predictions by up to 27% by focusing model training on the six loops responsible for antigen recognition rather than treating antibodies as generic proteins.
Novo Nordisk and AWS announced a sweeping strategic partnership on August 10, 2026, deploying Amazon Bio Discovery and Bedrock AgentCore through a new London co-innovation hub to accelerate drug candidate design, though experts caution clinical trials remain pharma's real bottleneck.
Insilico Medicine's AI-designed lung disease drug rentosertib won FDA Fast Track status in late July and is heading into Phase III, a test case for whether generative AI drug discovery can produce its first real approval.
A new AI model reads a routine biopsy slide and predicts cancer subtype, genetic mutations and survival outcomes across 32 solid tumor types at once, potentially shortcutting the wait for separate genomic testing.
UpDoc's FDA-cleared V1.0 platform uses a large language model to talk with type 2 diabetes patients and adjust insulin doses within physician-approved bounds, and it is now deployed at Cleveland Clinic, AHN, and UCSF, raising questions about how far generative AI should go in direct patient care.
RadNet's DeepHealth won FDA clearance for an AI system that standardizes breast ultrasound reads, cutting radiologist interpretation time by 37% and boosting cancer-detection sensitivity by 8% in clinical validation testing.