Show summary Hide summary
California lawmakers are weighing new limits on artificial intelligence in medicine as hospitals and tech vendors push AI tools into electronic health records. The debate — now centered on Assembly Bill 1979 — pits patient privacy and clinician judgment against potential efficiency gains from AI-assisted care.
Neela Paykel, a technology-focused counsel currently in residence at Mayer Brown and formerly general counsel at companies including Waymo and Hyperfine, says she relies on AI for routine work such as drafting emails and organizing schedules. She and other legal experts see the tools as practical assistants rather than replacements for professional skills.
Still, the rapid spread of AI in clinical settings has prompted sharp concern from labor groups, privacy advocates and some lawmakers. Assemblymember Rob Bonta introduced AB 1979 as a follow-up to last year’s legislation — AB 489 — which barred AI systems from impersonating licensed clinicians. The new measure aims to draw clearer lines around how automated outputs can be used in real-world medical decisions.
Stitt encampment sweeps displace residents: homeless camps keep reappearing
San Luis Obispo clerk-recorder: three candidates vie for control of county records
What the bill would do
The proposal focuses on AI embedded in electronic health records, specifically systems that provide what the industry calls clinical decision support. As amended in late April, AB 1979 would require health care professionals and organizations to exercise “independent professional judgment” before acting on recommendations generated by these systems.
The bill also seeks to bring so-called “health care chatbots” under existing privacy law by applying the Confidentiality of Medical Information Act to any business offering tools that let consumers upload or manage sensitive medical information.
Why privacy advocates and unions are pushing back
Opponents warn that linking patient records to large language models or third‑party services could expose data to analysis, storage or resale outside traditional medical privacy protections. The California Nurses Association has argued that uploading a medical record to a consumer-facing AI service can leave that information vulnerable to reuse for purposes unrelated to patient care.
Organizations recorded as opposing the bill in official analyses include Adventist Health, the Advanced Medical Technology Association, the California Hospital Association, Kaiser Permanente and the Civil Justice Association of California. In contrast, supporters listed by the Assembly include the California Labor Federation, the California Nurses Association, Consumer Watchdog and the California Peer Watch.
Potential benefits and limitations of AI in care
Proponents say properly designed decision-support tools can reduce medication errors, surface interacting medications, and ease cognitive load for clinicians juggling large amounts of patient data. Some studies cited by advocates suggest that integrating large language models into emergency department workflows could improve patient outcomes without reducing care quality.
Jill Speece, a professor of industrial and manufacturing engineering at Cal Poly with prior health-care consulting experience, says banning AI outright would be counterproductive. Instead, she argues, the focus should be on governance and safe deployment.
At the same time, students training for clinical careers express caution. Annalise Mui, a public health major who plans to become a physician assistant, welcomes AI as a task-oriented aid but warns against eroding the patient–clinician relationship or becoming overly dependent on automated outputs. She also notes that existing electronic health records already include privacy safeguards that any new tools must respect.
Key implications if AB 1979 passes
- For clinicians: Required documentation that clinical decisions were made independent of AI outputs could change workflows and liability considerations.
- For patients: Tighter rules aim to reduce the risk that personal health data will be reused by nonmedical vendors.
- For vendors: Companies supplying AI plugins or chatbots may face stricter compliance obligations under California privacy law.
- For policymakers: The bill could set a precedent for how states regulate AI in health systems nationwide.
Assemblymember Bonta has framed the bill as a safeguard: technology should support clinical judgment, not substitute for the human elements of care such as empathy and context-based decisions. Whether that balance will satisfy both innovators and privacy advocates remains to be seen.
AB 1979 is currently under consideration in the California Assembly. If approved by the Legislature and signed by the governor, its provisions would take effect on Jan. 1 of the following year.
