Part of the AI Regulation News hub.
California's AB 1979 and SB 503, signed on September 30, 2026 according to the Governor's office, would make hospitals, clinics and physician offices take reasonable steps to protect clinicians' independent judgment over clinical decision support AI, and would put bias duties on the developers and deployers of those systems
Neither bill puts the duty on the individual physician. Once operative, AB 1979 would aim at the facility or office that runs the AI, and SB 503 would ask for reasonable efforts against bias, with no promise of a bias-free system and an express rule that compliance is no defense to a discrimination claim.
Bottom line: Both are signed state statutes, according to the Governor's September 30, 2026 release, so they are binding law rather than guidance. The enrolled texts we read print no operative date, and we do not supply one. Until the duties apply, everything below describes what the sections would require.
Who this affects: California physicians and other licensed clinicians, medical directors and clinical leaders at health facilities, clinics, physician offices and group practices, health-system counsel and compliance officers, and companies that design or sell clinical decision support systems for use in California.
Issue date: Signed September 30, 2026, per the Governor's release. AB 1979 was enrolled September 1, 2026; SB 503 was enrolled August 27, 2026. Neither enrolled text states an effective or operative date.
What changed: AB 1979 adds Business and Professions Code section 22758.5, which once operative would put a reasonable-steps duty on health facilities, clinics, physician's offices and offices of a group practice to keep clinician judgment intact when clinical decision support informs care, plus a bar on AI independently performing licensed clinical functions. SB 503 adds section 22758, which once operative would set reasonable-efforts bias duties for developers and deployers of clinical decision support systems. Neither enrolled text states an operative date.
Analysis: On our reading, the practical exposure sits with the institution, not the bedside clinician, and SB 503 turns on documentation that developers would hand to deployers. A deployer that never asks for it, and never monitors, would struggle to show reasonable and proportionate steps once the section applies.
Primary sources: AB 1979 bill text, enrolled September 1, 2026 (California Legislative Information) · SB 503 bill text, enrolled August 27, 2026 (California Legislative Information) · Office of the Governor release listing the bills signed, September 30, 2026
- Instrument (EN)
- AB 1979 (Bonta), Health care services: artificial intelligence, adding Chapter 25.5, AI Clinical Decision Support (Bus. and Prof. Code s. 22758.5) and amending Civil Code ss. 56.05 and 56.06; SB 503 (Weber Pierson), Health care services: artificial intelligence, adding Chapter 25.3, Clinical Decision Support Systems (Bus. and Prof. Code s. 22758)
- Authority
- California Legislature; signed by Governor Gavin Newsom, according to the Office of the Governor release of September 30, 2026
- Jurisdiction
- California, United States
- Status
- Signed September 30, 2026, per the Governor's release. We read the enrolled texts, not chaptered versions, and did not verify chapter numbers
- Bindingness
- Statute, binding once operative. The enrolled texts state no operative date. SB 503 asks for reasonable efforts and reasonable and proportionate steps, not a guarantee of bias-free systems
- Issue date / next deadline
- Signed September 30, 2026. No compliance deadline or operative date appears in either enrolled text
- Enforcement named in the text
- Once operative, under the enrolled text of AB 1979, a violation by a physician would fall within the jurisdiction of the Medical Board of California or the Osteopathic Medical Board of California, and licensing boards could seek an injunction where a violation is unlicensed practice. SB 503: section 22758 names no penalty of its own
- Legislative path
- AB 1979: introduced February 13, 2026; passed Senate August 26 and Assembly August 27, 2026; enrolled September 1, 2026. SB 503: introduced February 19, 2025; passed Assembly August 24 and Senate August 25, 2026; enrolled August 27, 2026
- Primary source
- https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202520260AB1979
What does AB 1979 ask of hospitals, clinics and physician offices?
AB 1979's central duty, once the section takes effect, would be a reasonable-steps duty aimed at the place of care. Section 22758.5(a), in the enrolled text the Governor's office says was signed on September 30, would apply to "A health facility, clinic, physician's office, or office of a group practice". It does not name the individual physician as the duty-holder. Once the section takes effect, those entities would have to take reasonable steps to ensure that a licensed health care provider, acting within their scope of practice, keeps the ability to exercise independent professional judgment whenever a patient's care is informed by the output of a clinical decision support system.
The enrolled text defines a clinical decision support system as an AI system that "produces a prediction, classification, recommendation, evaluation, or analysis that aids clinical decisionmaking related to timing of care, diagnosis, or treatment." Once section 22758.5 is operative, under that definition appointment booking, reminders, patient education, pre-visit materials and payment processing would fall outside it, to the extent the system's independent performance of those tasks does not require a professional license.
One drafting detail is worth a careful reader's time. The Legislative Counsel's Digest describes the protected person as a "licensed health care professional", but the operative subdivision says "licensed health care provider", which the section defines as a person licensed or certified under Division 2 of the Business and Professions Code. The operative wording is the one that would apply.
AB 1979 also reaches past clinical decision support. Once operative, its amendment to Civil Code section 56.06 would mean that a business offering a "health care chatbot", as newly defined in section 56.05, to consumers for managing their information or for diagnosis, treatment or management of a condition is deemed a provider of health care under the Confidentiality of Medical Information Act. The digest says this would expand the scope of a crime, since certain CMIA violations are misdemeanors under existing law.
Where does AB 1979 draw the line on AI doing licensed work?
The enrolled text draws the line at independent performance. Under section 22758.5(b)(1), once operative, the same facilities and offices would be barred from using or deploying a tool, system or device that includes AI to "Independently perform any clinical function that is required by law to be performed by a person with a professional license." A second limb covers AI that would direct, guide, supervise or instruct unlicensed personnel in performing such a function.
This is not a ban on clinical AI. The prohibition is tied to functions the law already reserves to licensed people, and to the AI acting independently or steering unlicensed staff. Once the section is operative, subdivision (b)(2) would leave room for a trainee to use such a tool as part of a supervised course of study or training program toward licensure, and subdivision (d) would exclude automated decision systems used for documentation and communication that do not involve professional judgment, giving as examples automated messages about health record updates, reminders, and helping patients find information they ask for.
Enforcement in subdivision (c) of the enrolled text would run through licensing once the section is operative. A violation of the section "by a physician" would fall within the jurisdiction of the Medical Board of California or the Osteopathic Medical Board of California, and where a violation amounts to practising a health care profession without a license, the relevant board could seek an injunction or restraining order under section 125.5. The text reads that way even though the duties in (a) and (b) would sit with facilities and offices. On our reading, that matters most where a physician owns or runs the office in question; the text does not explain the pairing, and nothing in it shields physicians from other liability. Subdivision (c)(3) would preserve any remedy otherwise available to a licensing board or enforcement agency.
What does SB 503 ask of clinical decision support developers and deployers?
Once the section applies, SB 503 would ask for reasonable efforts, documentation and monitoring. SB 503 adds section 22758 and uses the same definition of a clinical decision support system as AB 1979, almost word for word. Developers and deployers would make reasonable efforts to identify systems developed for use by deployers that are known, or have a reasonably foreseeable risk, of biased impacts when deployed in health programs or activities. A "biased impact" is defined as "an adverse impact, including diminished access to health care, quality of care, or outcomes, on an individual based on their protected characteristics", with protected characteristics taken from Civil Code section 51(b).
Once section 22758 applies, for identified systems, developers would make reasonable efforts to mitigate the risk, and would make available to deployers a statement of intended uses and known or reasonably foreseeable risks, together with documentation covering seven items: high-level summaries of training data types, including demographic representativeness where demographic data is available; how the system was evaluated for performance, limitations and bias mitigation; data governance measures; intended benefits and expected outputs; known or foreseeable bias risks and mitigation efforts; recommendations on use, monitoring and risk mitigation; and any other documentation reasonably necessary for deployers to understand outputs and monitor for biased impacts.
Once operative, that documentation would be due on request or at initial sale, whichever is earlier, and on material updates. Once operative, paragraph (4) would offer two routes "In complying with paragraph (3)": developers may adhere to nationally recognized or widely adopted industry standards developed through multistakeholder consensus and "relevant to bias testing and artificial intelligence systems in health care", or "provide the results from regularly conducted algorithmic impact assessments using industry-accepted methodologies". Those results would need to include information relevant to the documentation items in paragraph (3)(B) and be appropriate for the system's intended use, and either route would apply only to the extent such standards or assessments exist. Paragraph (7) would add a narrow carve-out: a deployer that also develops a system for internal use would not need to generate the paragraph (3)(B) documentation if information consistent with those documentation and transparency requirements is readily available to it. The carve-out would cover (3)(B) only; the paragraph (3)(A) statement of intended uses and known or reasonably foreseeable risks would not be exempted.
Deployers, defined as a health facility, clinic, physician's office or office of a group practice that uses a clinical decision support system, would regularly monitor identified systems and "take reasonable and proportionate steps to mitigate the known or reasonably foreseeable risk of biased impacts." Nothing in the text promises or demands a bias-free system. The standard is effort and proportion, and it applies only once the section takes effect.
How do the two bills fit together, and where do they differ?
The two bills share an audience, but their definitions differ. Both add chapters to Division 8 of the Business and Professions Code, AB 1979 as Chapter 25.5 and SB 503 as Chapter 25.3, and both use near-identical clinical decision support definitions. Yet AB 1979 takes its definition of artificial intelligence from Health and Safety Code section 1339.75, while SB 503 takes it from Government Code section 11546.45.5. AB 1979 borrows its definitions of physician's office and office of a group practice from section 1339.75; SB 503 writes its own. We have not compared those cross-referenced provisions, so we do not say whether the differences change who is covered.
Once operative, SB 503 would close one door explicitly. Subdivision (d) says the section adds to, and does not replace, other state law on AI or automated decision systems, and that "Compliance with this section shall not be used as a defense to a claim of unlawful discrimination." For health-system counsel, the practical point is that a section 22758 compliance file would not by itself answer a discrimination claim.
The Governor's summary is looser than the text. The release describes SB 503 as "Requiring developers of clinical decision tools to take reasonable steps to reduce known or predictable bias in how these systems are used." Once section 22758 is operative, the text would ask for "reasonable efforts" from developers, would put separate monitoring and "reasonable and proportionate steps" on deployers, and speaks of known or reasonably foreseeable risk. Where the two differ, the statute governs.
When do the duties apply?
The documents we read do not answer that question. Neither enrolled text contains an effective or operative date clause, and the Governor's release gives none. We have not checked chaptering records or any general rule on when California statutes take effect, so this page states no start date.
The dates the documents do print are these. AB 1979 was introduced February 13, 2026, passed the Senate on August 26 and the Assembly on August 27, and was enrolled on September 1, 2026. SB 503 was introduced February 19, 2025, passed the Assembly on August 24 and the Senate on August 25, 2026, and was enrolled on August 27, 2026. The Office of the Governor lists both among bills signed on September 30, 2026.
Until a start date is confirmed, the duties described on this page are best treated as what the law would require once operative. Facilities, clinicians and vendors already face existing law, including the patient-communication disclaimer rule the AB 1979 and SB 503 digests describe as current law.
What we did not verify
What we opened: the California Legislative Information bill-text pages for AB 1979 (version enrolled September 1, 2026) and SB 503 (version enrolled August 27, 2026), including each Legislative Counsel's Digest, and the Office of the Governor release dated September 30, 2026, which lists both bills among legislation signed that day.
What we did not open: the chaptered versions and chapter numbers of either bill, any signing message for AB 1979 or SB 503 (the release links signing messages for other bills, not these two), committee analyses, vote records, and the cross-referenced provisions, including Health and Safety Code sections 1339.75 and 1250, Government Code section 11546.45.5, Civil Code section 51 and Business and Professions Code section 125.5. We did not review any other bill signed or acted on that day.
What we refuse to claim: an effective or operative date for either bill; that clinical AI is banned in California; that physicians are shielded from malpractice or any other liability; that SB 503 requires bias-free systems; that either bill creates a private right of action, which neither section we read mentions; or that these are the first laws of their kind. The release's characterisations are the Governor's office's own.
Quotations from the bills are copied from the enrolled text, with typographic apostrophes and quotation marks folded to straight ASCII.
Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.
California health facilities, clinics and physician offices that use clinical decision support AI would need a demonstrable way to keep the clinician's independent judgment in the loop, and to keep AI from independently doing licensed work, once AB 1979 applies. Once SB 503 applies, deployers and vendors should expect its documentation to move with the contract and to be monitored against, while remembering that compliance with section 22758 would not be a defense to a discrimination claim. Confirm the operative date from the chaptered statute before setting deadlines.
Source File
https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202520260AB1979
Open the AB 1979 and SB 503 bill-text pages on leginfo.legislature.ca.gov and confirm: the enrolled versions dated September 1, 2026 and August 27, 2026; section 22758.5(a), (b), (c) and (d) in AB 1979; section 22758(a)(1) to (8) and (d) in SB 503; and the absence of an operative-date clause. Then open the Governor's September 30, 2026 release and confirm both bills appear in the list of legislation signed.
A health facility, clinic, physician's office, or office of a group practice shall take reasonable steps to ensure that a licensed health care provider, acting within their scope of practice, retains the ability to exercise independent professional judgment in their care of a patient whenever that care is informed by the output of a clinical decision support system. ยท AB 1979, Business and Professions Code s. 22758.5(a), enrolled September 1, 2026; signed September 30, 2026 per the Governor's office
FAQ
Does AB 1979 ban AI in clinical care?
No. Once operative, it would bar health facilities, clinics, physician's offices and offices of a group practice from using AI to independently perform clinical functions that the law reserves to licensed people, or to direct unlicensed staff in those functions. Nothing in section 22758.5 bars clinical decision support that informs a clinician's care; for that use the section sets the reasonable-steps duty on judgment.
Who carries the AB 1979 duty, the physician or the practice?
The enrolled text would place the duties, once operative, on the health facility, clinic, physician's office or office of a group practice. A violation by a physician would then fall within the jurisdiction of the Medical Board of California or the Osteopathic Medical Board of California. The bill does not say physicians are shielded from liability, and neither enrolled text states an operative date.
Does SB 503 require clinical AI to be free of bias?
No. Once the section applies, developers would make reasonable efforts to identify and mitigate known or reasonably foreseeable bias risk and supply specified documentation, and deployers would monitor and take reasonable and proportionate mitigation steps. Compliance would expressly not be a defense to a claim of unlawful discrimination.
When do AB 1979 and SB 503 take effect?
The enrolled texts we read state no operative date, and the Governor's September 30, 2026 release gives none. We have not verified a start date and recommend confirming it from the chaptered statutes before planning compliance deadlines.
Related briefings
Sponsored Training
Practical AI training for regulated professionals, built around verification, documentation and a defensible process. See the courses.