The brief · 26 March 2026
Today healthcare is testing speed while finance is testing load: a nuclear case file that used to take weeks now wraps up in a day, and a Merrill advisor gets back four hours per client. Anthropic ships its biggest model yet, and the EU backs off the AI Act a little more.
Avery answers questions about coverage, claims, and appointments in plain language, around the clock. The open question: whether an insurer-built assistant nudges answers toward its own bottom line when you ask about your care.
practitioners › Launched March 26, 2026, available to all UnitedHealthcare members.
UnitedHealthcare →Repairing a mitral valve without opening the chest means tracking a tiny device in real time inside a beating heart. EchoNavigator R5.0 now overlays automatic device tracking onto the ultrasound image during the procedure. Less time spent interpreting the image, more time spent operating.
practitioners › FDA 510(k) clearance granted March 26, 2026, for DeviceGuide software built into EchoNavigator R5.0, covering minimally invasive mitral valve repair procedures.
Philips →25,000 financial advisors will have their meetings prepped, summarized, and followed up on by an AI that pulls from Salesforce and Zoom. In practice, that frees up time for actual advising, the part clients pay top dollar for. The open question is whether that time gets reinvested in better advice or just gets swallowed by a busier calendar.
practitioners › The tool draws on Salesforce and Zoom data to prep meeting materials, summarize conversations, and suggest follow-up actions, with a claimed savings of up to four hours per meeting.
Bank of America →A leak forced Anthropic's hand, confirming Mythos's existence ahead of schedule. The model is already circulating among a handful of vetted customers, showing marked gains in reasoning and coding. The bigger concern: its cybersecurity risk profile is rated higher than anything Anthropic has shipped before.
practitioners › Codenamed Capybara, access limited to early testers, no public release date announced.
Anthropic →The vote confirms what was flagged on March 18: the AI Act timeline is slipping again, this time pushing certain obligations to late 2027. The new element in the text is an explicit ban on apps that generate non-consensual nude images from real photos.
practitioners › Vote on March 26, 569 in favor, 45 against, 23 abstentions: high-risk obligations now pushed to December 2027 and August 2028.
European Parliament →