AMA Announces Framework for Safe Use of AI Agents in Healthcare

The American Medical Association released a governance framework to oversee agentic AI tools that can perform clinical tasks autonomously. The framework asks four questions: what function the AI will perform, the evidence for its performance, the conditions under which it may operate, and who is responsible for its outcomes. It stresses that AI solutions must be evidence‑based and maintain high quality before assisting patient care, keeping physicians central to development and deployment.

A recent survey of 250 U.S. healthcare leaders found that about 28 % already use agentic AI tools and 44 % are running pilots or proof‑of‑concept projects. However, 72 % say these tools are sometimes deployed without IT approval, raising concerns about compliance, regulatory violations, and unauthorized access to protected health information. The lack of oversight could lead to patient harm if an AI tool makes an inaccurate decision before a clinician intervenes. The survey highlights the need for clear governance.

Regulators at the state level have responded with a patchwork of laws. In 2025, 47 states introduced more than 250 AI bills affecting healthcare, according to legal firm Manatt. Some states tightened restrictions on AI‑based care, while others experimented with broader use. This variation makes it difficult for healthcare leaders to navigate the regulatory landscape while adopting AI technologies.