NHS rolls out Microsoft Copilot to 505,000 clinicians

1. ARPA-H selected innovation teams in June 2026 for the ADVOCATE program, a three-year federally funded effort to build the first FDA-authorized autonomous AI agent capable of providing around-the-clock specialty care for cardiovascular disease patients.

2. The program’s deeper significance is regulatory: ADVOCATE is designed to force the FDA to define a durable authorization pathway for AI agents that take real clinical actions, a category with no existing approved framework.

ARPA-H selected the teams for ADVOCATE in June 2026. The program, short for Agentic AI-Enabled Cardiovascular Care Transformation, funds a roughly three-year competition to build what does not yet exist: an FDA-authorized autonomous AI agent that can manage cardiovascular disease patients around the clock without a physician reviewing each decision in real time. The program is structured around two parallel systems. The clinical agent would handle the day-to-day work of chronic cardiovascular disease management: adjusting medication doses within pre-specified parameters, coordinating appointment scheduling, providing personalized dietary and exercise guidance based on patient-reported data and wearable device inputs, and flagging deterioration patterns that require human escalation. The supervisory agent would monitor the clinical agent continuously after deployment, watching for accuracy drift, unexpected outputs, or safety signals in a system that updates over time. That second component is actually the more technically and regulatorily novel of the two. Most cleared healthcare AI tools are validated once at the time of authorization and then deployed with minimal structured post-market monitoring. A cardiovascular AI agent that learns from new patient data will change over time, and having a separate AI system dedicated to detecting when that change becomes clinically unsafe is an architectural approach the field has not implemented at scale. As STAT News reported when the program launched, the transition from AI tools that give physicians information to AI agents that take actions on patients is the hardest problem in healthcare AI governance. Adjusting a medication dose is not a recommendation: it is a clinical action with direct patient consequences, and the liability and safety architecture required for that kind of authorization does not currently exist in FDA guidance. ADVOCATE is designed to build it. ARPA-H is working directly with FDA leadership throughout the program to develop the regulatory framework alongside the technology, with the explicit goal of having a first-of-its-kind clinical agentic AI authorization in hand by the time a winning team’s system is ready for deployment. For cardiologists and primary care physicians managing chronic heart failure, arrhythmia, and post-intervention patients, nothing changes at the bedside while this program runs. The relevance is prospective. Cardiovascular disease kills more Americans annually than any other condition, and continuous specialist-level monitoring outside the clinic is one of the most significant gaps in how high-risk patients are currently managed. Most patients with heart failure see their cardiologist a few times a year. An FDA-authorized AI agent capable of monitoring wearable data, adjusting diuretics, and escalating when needed could function as a continuous clinical presence that no current care model provides. Whether ADVOCATE produces a validated system on its three-year timeline, and whether that system can be deployed and reimbursed at the scale required to matter for that patient population, are the questions the program’s progress over the next several years will begin to answer.