Healthcare’s first AI registry launches soon with Kaiser Permanente, Stanford Medicine on board

The Coalition for Health AI (CHAI) is taking a significant step toward standardizing AI transparency in healthcare by creating a centralized registry for AI “nutrition labels”—detailed model cards that document an AI tool’s development, risks, and capabilities. This initiative, developed in partnership with IT consulting firm Avanade, aims to streamline communication between AI vendors and healthcare providers while making critical AI information accessible to the public, marking a crucial advancement in responsible AI deployment in healthcare.

The big picture: CHAI’s new registry will serve as a public repository for health AI model cards, building on their existing template system that has gained widespread industry adoption.

• The registry will be accessible to everyone, from healthcare professionals to the general public, with no membership requirements.
• All submissions will undergo human review to ensure information accuracy and trustworthiness.

Key partnerships: Several leading healthcare institutions have already signed the Model Card Call to Action, demonstrating strong industry support.

• Duke University Health System has become the first organization to join the registry.
• Cleveland Clinic, Kaiser Permanente, and Stanford Medicine have endorsed the initiative.
• More than a dozen AI solution providers have also pledged their support.

Behind the numbers: CHAI’s collaborative network includes nearly 3,000 healthcare organizations from both private and public sectors.

What they’re saying: “If everyone has a model card and they keep it to themselves, it’s actually not that valuable,” explained Brenton Hill, CHAI’s head of operations and general counsel.

• Michael Pencina, Duke University’s chief data scientist and CHAI board member, noted their pioneering role in “paving the way for additional members to utilize this important resource.”

Challenges ahead: The registry faces several technical and practical hurdles before reaching its full potential.

• The model card concept needs adaptation for generative AI, which requires frequent updates.
• Current PDF-based templates must be integrated into the registry and reformatted for better readability.
• The review timeline will vary based on submission volume.

What’s next: The registry will launch at the Healthcare Information and Management Systems Society Conference in Las Vegas, running from March 3-6.

• CHAI plans to develop integration capabilities with electronic health records.
• The organization aims to create flexible viewing options for different stakeholders, from executives to healthcare providers.