Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Health Equity topic
No spam. Unsubscribe anytime.
Safety-net providers and researchers warn AI could deepen disparities without data, funding and governance
Summary
Representatives of safety-net organizations, researchers and foundations told the Assembly committees that AI can help many Californians but will widen gaps unless Medi‑Cal and other public programs support data access, validation and sustained governance for under-resourced providers.
Speakers at the Joint Informational Hearing stressed that the state’s safety‑net providers face unique obstacles to safe AI adoption: limited data infrastructure, scarce funds for purchasing and sustaining tools, and insufficient capacity to run local validation and monitoring.
Cara Carter, senior vice president for programs and strategy at the California Health Care Foundation, described interviews with safety‑net providers and said many are eager to adopt AI but lack capital, staff time and technical readiness. Carter said CHCF is preparing listening activities to capture patients’ views and urged state support for data infrastructure, workforce training and governance.
Dr. Ziad Obermeyer, a physician and researcher at UC Berkeley, described prior research that found a widely used predictive algorithm prioritized patients who would generate high costs, not those who were medically high-risk — a disparity that left many Black patients without needed interventions. “Those algorithms predicted who was gonna generate high health care costs,” he said, and cautioned that automating flawed signals scales harm.
Representatives from community health‑IT networks and clinics told the committee that basic interoperability and up‑to‑date electronic health record environments are prerequisites for equitable AI. OCHIN’s public comment asked the state to invest in infrastructure, workforce training and streamlined regulations so community clinics and public‑health agencies can adopt AI safely and avoid an “AI chasm.”
Panelists proposed practical steps for policymakers: create privacy‑preserving mechanisms for aggregate data sharing, fund local validation and post‑deployment monitoring for safety‑net sites, and prioritize primary‑care use cases that directly serve underserved populations.
Panelists cautioned that policy choices could either narrow or widen gaps. Fawad Bhatt, founder of Penguin AI, advised starting with administrative automation (claims, prior authorization) that is structured and measurable before moving aggressively into higher‑risk clinical decision tools.
The committee is considering the testimony as it evaluates policy options for Medi‑Cal procurement, data sharing standards and targeted grants to ensure tools benefit rather than harm underserved Californians.
