Health care services: artificial intelligence

AI tools that influence the timing of care, diagnosis, or treatment can worsen access, quality, or outcomes for people with protected characteristics. The law creates transparency and ongoing risk-management duties for the companies and providers using those tools.

What the law does ​

• Requires developers and deployers to make reasonable efforts to identify clinical decision-support systems with known or reasonably foreseeable risks of biased impacts in health programs or activities.
• Requires developers to make reasonable efforts to mitigate those risks.
• Requires developers to give deployers intended-use and risk statements, plus documentation on training-data types and representativeness, testing, data governance, expected outputs, known bias risks, mitigation efforts, and recommended monitoring.
• Requires developers to provide required documentation upon request or initial sale, whichever comes first, and after material updates.
• Requires deployers to regularly monitor covered systems and take reasonable, proportionate steps to mitigate known or foreseeable biased impacts.
• Allows use of recognized health care AI bias-testing standards or algorithmic impact assessments where available.
• Excludes tools limited to administrative functions, such as appointment scheduling, reminders, patient education materials, and payment processing, when they do not require a professional license.
• Preserves other state AI and discrimination laws; compliance cannot be used as a defense to an unlawful-discrimination claim.

Who it affects ​

• Developers that design, code, substantially modify, or otherwise produce clinical decision-support systems for commercial or public use.
• Health facilities, clinics, physician offices, and group-practice offices that use covered systems.
• Patients whose care access, quality, or outcomes could be affected by biased AI-supported clinical decisions.

Context ​

A clinical decision-support system is an AI system that produces predictions, classifications, recommendations, evaluations, or analyses to aid clinical decisions about care timing, diagnosis, or treatment.