Loading page…
AB 1979 takes effect January 1, 2027. It preserves AI-assisted care while requiring health facilities to protect professional judgment.
Listen to this story
For California healthcare providers, AB 1979 turns the distinction between AI support and AI-led clinical action into a compliance requirement: facilities and practices must preserve clinicians’ independent judgment, while AI cannot perform licensed tasks such as prescribing or ordering tests. The law takes effect January 1, 2027, and licensing boards can seek court orders over violations. Which existing workflows cross the line remains unresolved, leaving providers to assess their AI deployments before the effective date.
The law covers AI predictions, classifications, recommendations, evaluations, and analyses that inform decisions about care timing, diagnosis, or treatment.
Documentation, communication, and some automated patient messages remain permitted; the measure does not ban AI use in medicine.
The California Medical Association shifted from opposing earlier bill language to neutrality after its scope was narrowed; the California Nurses Association sponsored it.
California’s new healthcare AI law allows software to inform treatment, but not independently perform clinical work that legally requires a licensed professional. Gov. Gavin Newsom signed AB 1979 on September 30, 2026. Starting January 1, 2027, covered healthcare settings must also take reasonable steps to preserve clinicians’ ability to exercise their own professional judgment.
The law addresses two different uses of AI. One is a tool that helps a clinician decide what to do. The other independently performs a function reserved by law for a licensed person. AB 1979 permits the first while prohibiting the second, including tasks such as prescribing medication or ordering tests.
For AI-assisted decisions, the obligation falls on health facilities, clinics, physicians’ offices and group practices. They must take reasonable steps to ensure licensed providers retain independent judgment while acting within their professional scope. The requirement concerns the setting in which care is delivered, not just the individual clinician consulting a tool.
The covered technology is defined by what it contributes to care. A clinical decision-support system is an AI system that produces a prediction, classification, recommendation, evaluation or analysis to aid clinical decisions. Those decisions can concern when care happens, what condition a patient has or how it should be treated.
That definition reaches beyond tools offering a diagnosis or a treatment recommendation. Predictions and classifications that inform the timing of care also fall within it. The professional-judgment requirement therefore applies to several forms of AI output, rather than only to software that gives an explicit instruction about a patient’s treatment.
Protecting patient safety, keeping our professional workforce engaged, and preserving the integrity of health care requires that we not rely exclusively on AI to do things quicker and cheaper.
Assemblymember Mia Bonta, who carried AB 1979, in a statement
AB 1979 is not a blanket prohibition on AI in medicine. Beyond allowing clinicians to consult AI when making care decisions, it preserves uses for documentation and communication, including some automated messages to patients. These permitted uses sit alongside the prohibition on independently performing clinical functions that require a professional license.
The distinction followed changes to the bill. The California Medical Association opposed earlier versions, arguing that they would broadly prohibit AI tools in medicine. After Bonta narrowed the language, the physicians’ group was neutral on the law. The California Hospital Association also took a neutral position, while the California Nurses Association sponsored the measure.
Hospital association spokesperson David Simon said clinicians should retain ultimate authority over patient-care decisions and AI should support them. His position preserved a role for tools that benefit patients while backing the principle that physicians and nurses remain in charge of care.
Licensing boards are authorized to pursue injunctions or restraining orders over violations. The law thus gives boards a route to seek orders against prohibited conduct, alongside its requirements for facilities to preserve clinicians’ judgment.
Which existing hospital or clinic practices might violate the new boundary remains unclear. In The Mercury News’s October 10 coverage, spokespeople for Bonta and the California Nurses Association did not identify examples of current AI uses that the law would prohibit.
The law could also affect disputes after care goes wrong. Henry Norwood, a healthcare attorney at defense firm Kaufman Dolowich, expects more malpractice suits alleging that doctors relied too heavily on AI guidance, especially in cases involving improperly delivered care. That is a forecast about litigation, not an established outcome of the new law.
Loading discussion...
Join the conversation
Explain whether a legal obligation is enough or stronger checks are needed.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.