New California laws will strengthen independent medical judgment in 2027. A Los Angeles hospital’s AI oversight process shows how the debate is moving from legislation into patient care.
Artificial intelligence can help doctors interpret medical information, identify possible diagnoses and evaluate treatments. But when an AI system recommends a course of action, who decides whether it is right for the patient?
California is preparing to strengthen its answer.
Beginning January 1, 2027, a new state law will require covered hospitals, clinics and medical practices to take reasonable steps to preserve licensed clinicians’ independent judgment when AI informs medical care.
The question is already relevant in Los Angeles, where Cedars-Sinai has deployed AI tools to assist clinical decisions and publicly described a process for reviewing them.
What does California’s new AI law require?
Assembly Bill 1979, signed in September 2026, addresses AI systems that support decisions about diagnosis, treatment or the timing of care.
It requires covered facilities and practices to preserve licensed professionals’ ability to exercise independent judgment.
The law also prohibits AI systems from independently performing clinical functions that legally require a professional license or directing unlicensed workers to perform those functions, subject to a supervised-trainee exception.
The protection is not that AI can never influence a medical decision. It is that professional judgment must remain independent and clinical functions requiring licensure cannot be delegated to AI.
The law does not guarantee that every AI recommendation is accurate or require separate human approval for every suggestion.
How is one Los Angeles hospital overseeing AI?
In May 2026, Cedars-Sinai announced that its clinicians could use OpenEvidence, an AI-enabled clinical reference system integrated with patient records.
The system allows physicians and other health professionals to research medical questions using published evidence alongside a patient’s medical history, medications, allergies and other relevant information.
Cedars-Sinai also described an institutional oversight process.
According to the hospital, AI tools undergo review by a designated committee before deployment. The committee includes clinical experts, data scientists, administrators and other specialists.
Its review addresses human oversight, verification of AI-generated information, privacy and potential effects on care.
This is a concrete example of a hospital describing how it governs clinical AI.
But the public announcement does not reveal how often clinicians reject AI recommendations, how individual decisions are documented or what independent evaluations have found.
Those remain questions for further reporting.
Why does human oversight matter to patients?
A 2026 Pew Research Center survey found that 60% of U.S. adults would feel uncomfortable with their doctor using AI for diagnoses or treatments.
Yet 60% were open to AI analyzing medical scans when a human performed the final review, compared with 6% who supported AI doing so independently.
The findings suggest that many patients distinguish between AI assisting clinicians and AI acting alone.
The American Medical Association’s October governance framework similarly asks what evidence supports a clinical AI system, what oversight is appropriate and who is accountable for outcomes.
That framework is voluntary guidance, not a binding law.
Will patients know when AI is involved?
Not necessarily.
California’s existing AB 3030 requires notices for certain AI-generated patient communications about clinical information.
However, the disclosure requirement does not apply when a licensed or certified health professional reads and reviews the communication.
The new AB 1979 does not create a universal requirement to notify patients whenever clinical AI is used.
Patients can ask whether AI influenced their diagnosis, treatment or medical communications and whether a clinician reviewed the information.
What about biased AI recommendations?
A second California law, SB 503, takes effect in January 2027.
It requires developers and health care organizations using covered clinical decision-support systems to make reasonable efforts to identify foreseeable biased impacts.
Developers must address identified risks and provide relevant information about systems’ intended uses, limitations and evaluation. Organizations deploying covered tools must monitor them and take reasonable, proportionate steps to mitigate known or foreseeable biased impacts.
The requirements do not guarantee that every AI tool will perform equally well for every patient.
Can AI deny insurance coverage or provide therapy?
Insurance coverage decisions follow different rules.
California’s SB 1120, effective in 2025, restricts AI use in insurance utilization reviews and preserves qualified human decision-making for medical-necessity denials, delays or modifications.
An insurer deciding whether to pay for care is not the same as a clinician deciding what treatment a patient needs.
Mental-health chatbots and AI-assisted therapy also raise separate questions involving professional licensing, privacy and patient safety. Their legal requirements should not be confused with hospital clinical decision support.
What does this mean for Latino patients?
Pew found that 41% of Hispanic adults expected AI to worsen patient health outcomes, compared with 31% expecting improvements.
Those figures measure public expectations, not actual medical outcomes.
For California families who prefer Spanish-language care, meaningful oversight also raises questions about accurate communication, language access and whether AI systems have been evaluated for diverse patient populations.
Parriva previously examined AI’s potential benefits and risks for immigrant families seeking health care.
What should patients ask their doctors?
Patients can ask whether AI influenced a recommendation, what clinical evidence supports the decision and whether a licensed professional independently evaluated the information.
They can also request language assistance, clarification or information about obtaining a second opinion.
A recent KFF analysis identifies human oversight as a growing priority in health care AI policy.
California has established new legal safeguards, and Cedars-Sinai has described one institutional approach.
The next accountability test is whether hospitals can demonstrate that their safeguards work when real patients’ medical decisions are at stake.
