New York's Office for the Aging says its ElliQ companion-robot pilot cut loneliness among 800-plus seniors by 95%, with users engaging more than 30 times a day. The figure comes entirely from state and vendor data rather than an independent study, even as the program heads into a second year.
Healthleap, a startup that uses AI to mine hospital notes and flag malnourished and otherwise undiagnosed patients, has raised $38 million from Sequoia Capital, First Round Capital, and Hummingbird Ventures, as it expands from 50 hospitals toward a roadmap of more than 40 conditions.
Clairity Breast, an FDA-authorized AI tool that reads mammograms to predict five-year breast cancer risk, launched nationwide direct-to-consumer through Everlywell on October 1, 2026, bypassing the traditional physician-and-insurer pipeline. STAT News examined the move as part of its 'Paying for AI' series on the costs and tradeoffs of consumer-facing clinical AI.
Twelve major U.S. health systems, including Cedars-Sinai, Mount Sinai, and Houston Methodist, have formed the Diagnostic AI Consortium with Aidoc to standardize how AI-assisted diagnosis is deployed and governed amid a worsening radiologist shortage. The group plans to publish shared evidence and governance practices starting in 2027.
Recursion Pharmaceuticals says its AI platform designed REC-7735, a selective PI3Kα-mutant cancer inhibitor, from 242 candidate compounds to a cleared IND in 10 months; the drug is headed into a Phase 1/2 trial as the company leans on partnerships with Roche's Genentech and Sanofi.
Tempus AI has received FDA 510(k) clearance for ECG-MR, its fourth cleared ECG-AI product, which scans routine 12-lead heart tracings for signs of undiagnosed moderate-to-severe mitral regurgitation in at-risk patients 65 and older.
Massachusetts General Hospital and digital health company Dimagi are testing a 12-week AI chatbot meant to support patients in medication treatment for opioid use disorder between clinical visits, with a tightly screened group of 55 participants and no results yet published.
AI platforms like inFeedo's Amber are mining employees' email timestamps, meeting loads, and after-hours logins to flag burnout weeks before annual surveys would catch it, but the same data shows nearly 8 in 10 workers already check email after hours and worry about workplace surveillance.
A new AI model trained on 1,000 Reddit posts can flag the overlapping combination of body image distress, disordered eating, and metabolic symptoms common among women with PCOS, with researchers reaching 75.3% accuracy — while cautioning their own tool is fit only for screening, not diagnosis.
A federated-learning AI called OsteoPredict, presented at the World Congress on Osteoporosis 2026, found it can flag likely osteoporosis from routine chest X-rays with 91% sensitivity, potentially catching at-risk patients who never get a dedicated bone scan.