The Coalition for Health AI (CHAI), a non-profit that brings together health systems, technology companies, patient groups, and regulators, has published a set of governance playbooks to help health systems oversee their own use of artificial intelligence. The guidance was developed with more than 150 industry contributors and over 100 healthcare organisations, and is organised around eight dimensions of organisational risk.

Those dimensions cover an organisation's AI policy, governance structure, and resourcing; responsible management of the AI lifecycle; risk and impact assessments; data management; the oversight of third-party vendors; and education and training. Rather than prescribe particular products, the playbooks set out baseline controls that an organisation can adapt. CHAI says they are meant to scale from large academic medical centres to resource-constrained community clinics.

The release reflects a shift in health AI from asking whether a model performs well in testing to asking how an organisation governs it in use: who authorises deployment, how bias and drift are monitored, and what happens when a vendor changes a model. "The resulting playbooks are intentionally designed to democratize AI innovation," said Brian Anderson, CHAI's chief executive. The Joint Commission, which accredits US hospitals, supported the effort; its president, Jonathan Perlin, called the playbooks "a massive milestone."

The guidance is voluntary, and whether health systems adopt it, or have the staff to apply eight dimensions of governance in practice, will determine its effect. As a shared reference point, though, it gives organisations of very different sizes a common vocabulary for governing tools many are already deploying.