AMA agentic AI governance framework sets four questions

The American Medical Association (AMA) has released a governance framework for AI agents used in healthcare, MedTech Dive reported. The AMA agentic AI governance framework aims to keep their use safe and effective.

What the framework says

The framework rests on one principle. Agentic AI that performs clinical functions needs evidence behind it and must keep up high quality performance.

It sets out four questions about which tasks AI tools may perform, in which situations, and who is accountable for their actions. The AMA will use it to guide work with other organisations. The aim is for AI agents to behave safely and responsibly wherever their tasks reach patients.

The four questions

The AMA will ask these when it assesses clinical AI agents:

• Which clinical function is the AI performing?
• What evidence shows it can do that, and what follows if it breaks down?
• What may the AI do, and under what conditions?
• Who carries responsibility for the risks and outcomes?

The framework states that “evidence, authority, oversight and accountability must remain aligned in service of better patient care” as AI capabilities advance.

Why oversight matters

Agentic AI can work autonomously and finish complex tasks that generative AI cannot. Clinical teams already rely on it for tasks such as patient monitoring, diagnostics and ambient documentation.

Such tools can carry out multi step work unsupervised. A wrong or biased decision could therefore seriously harm a patient before any clinician steps in. Industry groups are weighing how to contain that risk. The AMA is stressing physicians' central role in building and deploying the technology.

Adoption and supervision gaps

A recent survey of 250 US healthcare leaders found about 28% had already deployed agentic AI tools. Another 44% were running pilots or proof of concept projects. A majority expect a “transformative” effect on clinical workflows.

Yet 72% said agentic tools are sometimes deployed without IT approval, pointing to a supervision gap. More than half listed possible compliance or regulatory violations, plus unauthorised access to protected health information, among their main worries.

Regulatory backdrop

Health AI is moving faster than regulation. With little substantive federal rule making, industry groups have formed to offer guidance and guardrails.

States have produced a patchwork. Some are restricting the technology, while others are trialling AI based care. Legal and consulting firm Manatt counts 47 states that introduced over 250 AI bills affecting healthcare in 2025.

Why it matters

The AMA agentic AI governance framework hands hospitals and vendors a set of questions to answer before AI agents take on clinical tasks. Medigear.uk follows the AMA agentic AI governance framework and wider healthcare AI news.

Source: MedTech Dive