Heart Association Issues Urgent AI Safety Guidelines for Healthcare

The American Heart Association has issued a groundbreaking science advisory calling for urgent action to address the rapid proliferation of artificial intelligence tools in healthcare that lack proper evaluation and monitoring. Published in the Association's flagship journal Circulation, the advisory titled "Pragmatic Approaches to the Evaluation and Monitoring of Artificial Intelligence in Healthcare" reveals a critical gap: while hundreds of health care AI tools have received FDA clearance, only a fraction undergo rigorous assessment for clinical impact, fairness, or bias. The U.S. Food and Drug Administration's review process covers just a small portion of the AI tools actually being developed and deployed in clinical settings, creating significant patient safety concerns.

The advisory introduces a comprehensive, risk-based framework built on four guiding principles to help health systems establish effective AI governance. These principles include strategic alignment, ethical evaluation, usefulness and effectiveness, and financial performance. Dr. Sneha S. Jain, volunteer vice chair for the writing group and Director of the GUIDE-AI Lab at Stanford Health Care, emphasized that "AI is transforming health care faster than traditional evaluation frameworks can keep up." The guidance aims to ensure AI tools deliver measurable clinical benefit while protecting patients from both known and unknown harms. A recent survey highlighted the urgency, finding that only 61% of hospitals using predictive AI tools validated them on local data before deployment, with fewer than half testing for bias.

The American Heart Association's extensive network through its Get With The Guidelines quality improvement programs, involving nearly 3,000 hospitals including over 500 rural facilities, positions the organization as a trusted leader in advancing responsible AI governance. The Association has committed substantial research funding exceeding $12 million in 2025 to test novel health care AI delivery strategies. The advisory stresses that monitoring cannot end after deployment, as AI tool performance may drift with changing clinical practices or patient populations. Dr. Lee H. Schwamm, writing group chair, stated that "responsible AI use is not optional, it's essential," underscoring the need for practical steps to ensure AI tools improve patient outcomes and support equitable, high-quality care across all healthcare settings.