Google on Wednesday shared a framework it says is designed to evaluate whether artificial intelligence systems for health are safe and helpful. The company said the whitepaper details an end-to-end approach for assessing AI use cases that address people’s health needs.
The company said the framework measures AI models in three areas: accuracy, safety, and usefulness. It said the approach combines closed-question benchmarks, autoraters, and expert human review to test common and high‑risk health scenarios. Google said the goal is to help determine if models provide useful, accurate responses that promote safe interactions in health contexts.
Google said it built and continues to use this framework to assess its own models for health uses. The company said it is sharing the methodology to promote transparency in how it evaluates AI for health. It also said it aims to support developers, researchers, and clinicians across the health ecosystem in building AI tools people can trust.
The company said the whitepaper was developed in partnership with health experts Dr. Elizabeth McGlynn and Dr. Dana Gelb Safran. It did not provide additional details on specific tests, datasets, or performance results.
Google did not say when or how external groups could access or adopt the framework beyond the publication of the whitepaper. The company did not specify which AI models or products it has evaluated with this approach.