FDA clears sharper EKG heart-attack detector as US funds autonomous cardiac AI
A rare FDA authorization for an AI that spots hard-to-catch heart attacks lands alongside a $62.7 million federal bet on AI agents that could manage heart failure patients on their own.
What happened: Two cardiac-AI developments landed within days of each other. The FDA granted a rare de novo authorization, a pathway used fewer than 10 times a year, to an algorithm from Powerful Medical that reads a standard 10-second EKG to help flag heart attacks that are easy to miss, aiming to route chest-pain patients to the right care faster. Separately, federal research agency ARPA-H announced ADVOCATE, a 62.7 million dollar, four-year program to build what it calls the first FDA-authorized agentic AI system for cardiovascular care, software that could act semi-independently to help manage heart failure patients between doctor visits, with 33.7 million dollars allocated for year one alone.
Why it matters: The most dangerous heart attacks happen when a coronary artery is fully blocked; every minute without treatment costs heart muscle, and doctors already scan EKGs for a telltale signature to rush patients to a catheterization lab. This tool targets cases where that signature is subtler and easy for a human reader to miss. The bigger federal bet targets a stubborn problem: heart failure and preventable cardiovascular disease keep killing people not because the medicine doesn't work, but because many patients, especially in rural or underserved areas, can't reliably reach a cardiologist.
How it works, plainly: Powerful Medical's tool applies pattern recognition to the same EKG waveform doctors already read, hunting for heart attacks that don't show textbook signs. ADVOCATE splits a harder problem into layers: Atman Health, Tempus AI, and Updoc are separately building patient-facing agents that could interact with heart failure patients between visits; Stanford is building a supervisory AI to watch those agents for unsafe or unusual recommendations; and Duke University with Kaiser Permanente will test real-world deployment across health systems. Johns Hopkins' Applied Physics Lab is the independent referee checking whether any of it actually works safely.
The rollout: Powerful Medical's clearance is done, so its tool can now legally reach the market, though full performance details sit behind a paywall. ADVOCATE is much earlier stage: the three companies building patient-facing agents must each submit an FDA-authorization package within 24 months, and it is unclear how many will make it that far. ARPA-H projects 28 billion dollars in annual savings for heart failure patients but hasn't published how that figure was calculated. No agentic AI system has ever been FDA-authorized for direct cardiovascular care, so either program could set new regulatory ground.