79% of hospital AI comes from EHR vendors. Here's why that's a problem

Context 2026 A 2025 national analysis revealed that 79% of AI models used in U.S. hospitals come from their EHR vendor. Health systems are choosing convenience over capability — and many do not even realize it. In this session from Context 2026, Ruth Reader, Editor at Second Opinion Media, leads a candid conversation with Julia Adler-Milstein and Lisa Bari, VP of Policy and Partnerships at Innovaccer, on the regulatory and structural forces shaping autonomous AI adoption in healthcare. In this video: → Why structural lock-in keeps health systems from accessing best-in-class AI → The state of information blocking enforcement and why it still lacks teeth → FHIR API performance gaps and what better policy looks like → Deep integration challenges for autonomous AI across fragmented data systems → How agentic AI changes the stakes for data access and patient safety Chapters: 00:00 The 79% statistic and what it means 01:52 Why health systems default to EHR-bundled AI 04:20 Structural lock-in and the stay-in-suite problem 06:27 FHIR APIs — policy headed right, execution falling short 10:22 What would make this policy better 12:45 Information blocking enforcement — where things stand 21:38 Enforcing the laws already on the books 27:07 Sandboxes and developer access 31:27 Balancing AI regulation with innovation 34:20 Agentic AI and the data access imperative 35:27 Audience Q&A — data ownership, ambient AI evaluation #HealthcareAI #HealthcareAutonomy #Context2026