The FDA's Digital Health Advisory Committee is weighing how to regulate AI therapy chatbots after Woebot's 2025 shutdown exposed the gap, while patient-safety group ECRI has named the chatbots the top health technology hazard of 2026.
Oura filed to go public on September 4, 2026, targeting a valuation above $16 billion and citing nearly 42 billion hours of physiological data behind its AI health models, even as a class-action lawsuit challenges the accuracy of its sleep-stage tracking.
Whoop is adding on-demand video consultations with licensed clinicians and a suite of generative-AI features, including a personal-context hub called My Memory and proactive coaching check-ins, to its fitness membership platform starting this summer.
DeepHealth has won FDA clearance for two new Breast Suite features: an AI tool that flags cardiovascular risk from routine screening mammograms with over 90% sensitivity, and an upgraded Mammo Dx that compares current and prior exams to cut unnecessary recalls.
A Mass General Brigham study of 21 AI models found they fail to produce an appropriate differential diagnosis more than 80% of the time, even though the same models reach the correct final diagnosis over 90% of the time once all clinical facts are already provided.
Medtronic has won FDA clearance for Stealth AXiS, a platform that combines AI-based surgical planning, real-time navigation and robotics for spine surgery, with a follow-on clearance expanding it into cranial and ENT procedures.
Revolve Surgical has won FDA 510(k) clearance for a zero-footprint surgical robot that mounts directly to the operating table instead of the floor, a design the Y Combinator-backed startup says is a first step toward "physical AI" in the operating room.
Artera has won FDA clearance for ArteraAI Breast, a same-day digital pathology AI tool that stratifies chemotherapy risk for HR-positive, HER2-negative breast cancer patients using their own resection tissue rather than a separate genomic assay.
Meta has begun production of its custom Iris AI chip, built with Broadcom on TSMC's 3-nanometer process, as part of a plan to double its AI computing capacity to 14 gigawatts by 2027 and reduce its reliance on Nvidia GPUs for internal workloads.
Cerebras Systems and Compute Nordic Finland announced a phased 165-megawatt AI data center in Mikkeli, Finland, backed by seven-year service contracts and an estimated €1.0-1.7 billion regional investment, as AI infrastructure spending increasingly targets Nordic markets with cheap, renewable power.