AI Enters the Exam Room, But Healthcare's Old Plumbing Pushes Back

September 2026 was the month AI in health care stopped being a demo and started being an argument about money and infrastructure. ARPA-H put nearly $100 million behind AI-driven rare disease diagnostics, OpenEvidence kept shipping specialist agents for oncology and beyond, and an FDA-cleared sepsis detection system showed how machine learning can catch what tired clinicians miss. But a Peterson Health Technology Institute report landed like a splash of cold water, warning that fee-for-service payment could turn autonomous clinical AI into a cost driver rather than a cost saver. Two separate pieces made the same point from different angles: hospitals are bolting new AI tools onto systems built in the 1990s, and burnout isn't fading, it's mutating. Governments spent the month building pipes, not just algorithms. New Zealand advanced a $60 million radiology overhaul meant to unify imaging data nationally. Malaysia yanked its digital health timeline forward by seventeen years, backing the shift with $250 million. HHS sent $77 million toward substance use and mental health programs, while a vaccine guidance vacuum at the federal level pushed four major medical societies to issue their own flu, COVID and RSV recommendations for the first time in decades. The common thread was institutions filling gaps that federal agencies left open. Science delivered its own quieter victories. Teva's antibody blocked gluten-induced gut damage in a mid-stage celiac trial, offering the first real treatment prospect beyond total avoidance. The FDA approved Mimrylo for polycythemia vera, ending years of reliance on routine phlebotomy for that rare blood disorder. A Dublin startup's sealant aimed to cut pneumothorax rates after lung biopsies, a complication that hits up to a third of the 800,000 procedures performed annually in the U.S. Even quantum computing found a health angle, with Pasqal and True Nexus using neutral-atom hardware to model the protein gelation behind plant-based food texture. The business side of digital health kept consolidating. Neil Patel of Redesign Health described mergers accelerating as investors reward durability over hype, a shift from the pandemic-era gold rush toward something more disciplined. Judi Health expanded its clinical partner network, betting that integration, not proliferation, wins employer contracts. Genentech's Charlotte Owens pushed the industry to confront a blunter fact: more than half of prescriptions go unfilled, no matter how good the underlying science is.