Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Health Ai topic
No spam. Unsubscribe anytime.
Committee advances SB 503 to require bias identification and monitoring for AI used in clinical settings
Summary
SACRAMENTO — The Assembly Health Committee voted July 8 to send SB 503, an AI transparency‑and‑bias measure by Senator Weber Pearson, to the Privacy and Consumer Protection Committee as amended.
SACRAMENTO — The Assembly Health Committee voted July 8 to send SB 503, an AI transparency‑and‑bias measure by Senator Weber Pearson, to the Privacy and Consumer Protection Committee as amended. The bill requires health‑care AI tools that support clinical decision making or resource allocation to be identified, evaluated for biased impacts, mitigated where possible, and monitored after deployment.
“AI tools do not operate in a vacuum,” Weber Pearson told the committee, adding that every step of model development and deployment “can be embedded with and reflect existing societal biases.” The sponsor said health‑care decision support tools trained on non‑representative data can reinforce or worsen disparities unless there are proactive requirements to test and monitor them.
Clinical and industry witnesses testified in favor. Dr. Steven Wang, a radiologist with Southern California Permanente Medical Group and chair of the California Medical Association’s health IT subcommittee, called the bill “a measured and practical first approach” that would build clinicians’ trust in AI. “SB 503 would require AI developers and deployers to make reasonable efforts to mitigate biased impacts when AI tools are used in clinical decision making and health care resource allocation,” Wang told the committee.
The California Medical Association, Kaiser Permanente, the California Hospital Association and several specialty societies provided support statements or witnesses, saying clinicians need guardrails and transparency to use AI safely in patient care.
Committee members stressed the developer/deployer division in practice and asked the author to continue clarifying definitions and the scope of entities covered. The bill’s amendments, which the author accepted, refine the definition of "deployer" and align other language to upcoming privacy committee review.
No primary witnesses spoke in opposition in the hearing room; several health and safety groups and hospital organizations registered support. With amendments accepted as described in the hearing record, the committee voted to pass SB 503 as amended to the Privacy and Consumer Protection Committee.
Ending: Lawmakers said the bill aims to balance innovation with patient safety by ensuring both developers and health‑care deployers take responsibility for identifying and mitigating bias in clinical AI tools.
