AI Frontier Institute

Health care services: artificial intelligence.

View primary source →

California Senate Bill 503 (Chapter 857), signed September 30, 2026, adds Chapter 25.3 (commencing with Section 22758) to Division 8 of the Business and Professions Code to regulate artificial intelligence-based clinical decision support systems (CDSS) in health care. The law defines a 'clinical decision support system' as an AI system that produces predictions, classifications, recommendations, evaluations, or analyses aiding clinical decisionmaking related to timing of care, diagnosis, or treatment, excluding routine administrative tools like appointment booking or payment processing. Under Section 22758, both developers and deployers must make reasonable efforts to identify CDSS that are known or have a reasonably foreseeable risk of biased impacts—defined as adverse impacts on individuals based on protected characteristics, including diminished access, quality of care, or outcomes. Developers must mitigate such risks and provide deployers with: (A) a statement of intended uses and known or foreseeable risks; and (B) detailed documentation covering training data demographics, performance evaluation methods, data governance measures, intended benefits, expected outputs, known bias risks and mitigations, and monitoring recommendations (Section 22758(a)(3)). This documentation must be provided upon request or at initial sale, and upon material updates. Developers may satisfy documentation requirements by adhering to nationally recognized multistakeholder consensus standards for AI bias testing in health care or by providing algorithmic impact assessments using industry-accepted methodologies (Section 22758(a)(4)). Deployers—defined as health facilities, clinics, physician's offices, or group practices using a CDSS—must regularly monitor deployed systems and take reasonable, proportionate steps to mitigate biased impacts (Section 22758(a)(8)). An entity may be both a developer and a deployer. The bill explicitly states that compliance does not serve as a defense to unlawful discrimination claims and does not replace other applicable state AI or automated decision-system laws.

← Back to the tracker