Oracle Health's AI-powered patient portal, now generally available across the US, lets patients ask plain-language questions about labs, medications and diabetes management directly from their EHR, with citations and highlighting meant to keep AI answers traceable rather than opaque.
The OpenAI Foundation is putting $125 million into Public Data for Health, funding open scientific datasets at UNC's cancer center, OpenADMET and CTD Commons in a bet that better biological data, not just bigger models, is the real bottleneck for AI in medicine.
OpenEvidence, used daily by more than 40% of U.S. physicians, split its clinical AI chatbot into three named models for speed and depth while holding back a fourth, Darwin, which it says scored perfectly on the MedQA benchmark, citing dual-use safety risks.
Elucid closed a $55 million Series D on September 2, 2026, bringing total funding to about $185 million to expand its FDA-cleared Plaque-IQ AI, which reads coronary CT scans to classify dangerous heart plaque, plus a new FFR-CT tool awaiting FDA review.
The FDA's new TEMPO pilot lets four generative-AI health tools, from SonderMind, Limbic, Cadence and Dexcom, reach Medicare patients before they receive full marketing authorization, trading pre-market review for real-world outcomes data on chronic conditions like depression, hypertension and prediabetes.
Google and Character.AI settled five lawsuits, including the Sewell Setzer III wrongful-death case, alleging AI companion chatbots contributed to harm and death among teenagers.
MyFitnessPal launched an AI Nutrition Coach built on its Cal AI acquisition, just as a NUTRITION 2026 study found AI calorie-photo apps average a 345-calorie estimation error.
Apple delayed its AI health coach, Project Mulberry, at WWDC 2026 but is shipping AI sleep-sound detection and an Apple Watch Series 12 with 5-second heart-rate sampling.
Google is retiring the Fitbit brand for a $9.99/month Gemini-powered Health Coach launching May 19, 2026, alongside a new $99.99 Fitbit Air tracker.
A Zhengzhou Normal University study found that a ChatGPT-based depression screening tool, BDI-FS-GPT, achieved 89.3% sensitivity, outperforming the widely used PHQ-9 questionnaire, though researchers caution larger trials are still needed.