AI Regulation Tracker / Health coverage and utilization review
California Puts a Licensed Physician, Not AI, in Charge of Coverage Denials
Effective January 1, 2025. California SB 1120, the Physicians Make Decisions Act, requires that a denial, delay, or modification of care for reasons of medical necessity be made by a licensed physician or a competent licensed clinician, not by an AI tool or algorithm acting on its own.
California SB 1120, the Physicians Make Decisions Act, has been in force since January 1, 2025, and it targets a specific problem that has grown with AI in health coverage: automated denial of care. The concern was straightforward. Health plans were increasingly running utilization review, the process that approves or denies claims and prior authorizations, with AI and algorithms in the loop, and there was real fear that a piece of software was effectively deciding whether a patient got treatment. SB 1120 draws the line. AI can assist the review. It cannot make the medical-necessity denial.
The statute is explicit. In the Health and Safety Code, it provides that "No individual, other than a licensed physician or a licensed health care professional who is competent to evaluate the specific clinical issues involved in the health care services requested by the provider, may deny or modify requests for authorization of health care services for an enrollee for reasons of medical necessity." The parallel Insurance Code provision says a determination of medical necessity "shall be made only by a licensed physician or licensed health care professional competent to evaluate the specific clinical issues involved." An algorithm is not a licensed physician, and that is the point.
What can AI still do under SB 1120?
Quite a lot, actually, and it is important not to overstate the ban. AI can be used to support utilization review. It can flag, sort, summarize, and route. What it cannot do is issue the adverse medical-necessity determination on its own. There is also a data-integrity requirement that matters in practice. Any AI tool used in this process has to base its assessment on the enrollee's own medical or clinical history and the clinical circumstances presented by the requesting provider, not on a broad dataset applied to the individual. That provision is aimed squarely at the practice of denying care by group statistics rather than by the patient in front of you.
Why this reaches beyond California plans
SB 1120 formalizes something the whole industry is being pushed toward. When a large state fixes the medical-necessity decision to a licensed human and makes the AI auditable by regulators, national plans tend to standardize on the stricter rule rather than run two processes. It also lands in the middle of a broader wave. Several states have moved on AI in coverage decisions and prior authorization from different angles, and the common thread is that a qualified human has to own the denial. California went first and biggest, which is why counsel and compliance teams outside the state still need to know it.
What physicians and review operations should do
For medical directors and clinicians doing utilization review, the compliance posture is clear. Make sure a licensed clinician competent in the relevant specialty is the one signing every medical-necessity denial, delay, or modification, and that the sign-off is a real review, not a rubber stamp on an algorithm's output. Keep the AI in an assistive role and document where the human judgment enters. Confirm the tool is drawing on the individual enrollee's record. And be ready for oversight, because the DMHC and the CDI can audit AI systems used in these decisions for accuracy, reliability, and compliance. The safest operating assumption is that a regulator may one day ask you to show that a person, not a model, denied the care.
Questions professionals are asking
Does SB 1120 ban AI in utilization review?
No. AI can support the review, flag, sort, and summarize. It cannot issue the medical-necessity denial, delay, or modification on its own. That determination must be made by a licensed physician or a licensed health care professional competent in the specific clinical issues involved.
What data can the AI use to make its assessment?
The tool must base its determination on the enrollee's own medical or clinical history and the clinical circumstances presented by the requesting provider, plus relevant information in the enrollee's record. It cannot substitute a broad group dataset for the individual patient.
Who enforces SB 1120?
The California Department of Managed Health Care and the Department of Insurance. They can audit and review AI systems used in utilization review to check accuracy, reliability, and compliance with the requirement that a licensed clinician make the medical-necessity call.
Does this affect plans based outside California?
It binds plans and insurers operating in California, but its practical reach is wider. National operations often standardize on the stricter rule, and it sits alongside a broader wave of state action requiring a qualified human to own coverage denials and prior-authorization decisions.
RELATED BRIEFINGS
Browse the full AI Regulation News tracker
Informational analysis for working professionals, not legal advice. Confirm how any statute or requirement applies to your situation with qualified counsel licensed in the relevant jurisdiction.