|
Amended
IN
Assembly
July 02, 2026 |
|
Amended
IN
Assembly
June 16, 2026 |
|
Amended
IN
Senate
January 05, 2026 |
|
Amended
IN
Senate
May 01, 2025 |
|
Amended
IN
Senate
March 26, 2025 |
CALIFORNIA LEGISLATURE—
2025–2026 REGULAR SESSION
Senate Bill
No. 813
| Introduced by Senator McNerney (Coauthors: Assembly Members Bauer-Kahan and Lowenthal) |
February 21, 2025 |
An act to add Chapter 14 (commencing with Section 8898) to Division 1 of Title 2 of the Government Code, relating to artificial intelligence.
LEGISLATIVE COUNSEL'S DIGEST
SB 813, as amended, McNerney.
California Artificial Intelligence Standards and Safety Commission: artificial intelligence safety standards.
Existing law requires, on or before September 1, 2024, the Department of Technology, within the Government Operations Agency, to conduct, in coordination with other interagency bodies as it deems appropriate, a comprehensive inventory of all high-risk automated decision systems that have been proposed for use, development, or procurement by, or are being used, developed, or procured by, any state agency. Existing law requires the department to annually submit a report of that comprehensive inventory to the Assembly Committee on Privacy and Consumer Protection and the Senate Committee on Governmental Organization.
Existing law, the Transparency in Frontier Artificial Intelligence Act, among other things related to ensuring the safety of certain artificial intelligence models, requires a large frontier developer to write, implement, and clearly
and conspicuously publish on its internet website a frontier AI framework that applies to the large frontier developer’s frontier models and describes how the large frontier developer approaches, among other things, incorporating national standards, international standards, and industry-consensus best practices into its frontier AI framework.
This bill would require the Business and Consumer Services Agency to establish, on or before July 1, 2027, the California Artificial Intelligence Standards and Safety Commission and would provide for its membership, as specified. The bill would make this provision inoperative if AB 1709 of the 2025–26 Regular Session is enacted and the e-Safety Advisory Commission is established pursuant to that measure.
This bill would require either the California Artificial Intelligence Standards and Safety Commission or the e-Safety Advisory Commission, as specified, to take certain actions related
to the safety of artificial intelligence, including developing standards for the assessment of AI systems and models by AI auditors, developing procedures for determining whether to suspend or terminate the registration of an AI auditor, and developing criteria for determining whether an AI auditor qualifies as an independent verification organization, as specified.
This bill would make its operation contingent upon the enactment of AB 1405 of the 2025–26 Regular Session.
The people of the State of California do enact as follows:
SECTION 1.
The Legislature finds and declares all of the following:(a) California is the global leader in artificial intelligence innovation and is home to the most artificial intelligence companies, researchers, and developers in the world.
(b) Artificial intelligence technologies have the potential to drive economic growth, improve public services, advance scientific discovery, and enhance the quality of life for Californians.
(c) As artificial intelligence systems become increasingly capable and widely deployed, ensuring their safe,
secure, and trustworthy development and use is essential to protecting consumers, workers, businesses, and the public.
(d) California has enacted laws addressing specific risks associated with artificial intelligence.
(e) Existing laws provide important safeguards, but due to continuing evolution of the technology, continued collaboration among industry, government, academic institutions, labor organizations, consumer advocates, civil society organizations, and other stakeholders is necessary to address emerging risks and establish best practices for artificial intelligence governance.
(f) Consensus-based standards have played an important role in promoting safety, interoperability, accountability, and public trust
across numerous industries and technologies.
(g) The development of artificial intelligence safety standards benefits from the expertise of diverse stakeholders with technical, economic, legal, workforce, and public interest perspectives.
(h) Establishing structured working groups composed of experts and stakeholders can facilitate the development of practical, evidence-based standards that evolve alongside advances in artificial intelligence technology.
(i) This approach creates a scalable artificial intelligence governance framework.
(j) It is the intent of the Legislature that the standards and processes established by this act support the safe development and deployment
of artificial intelligence systems while preserving innovation, competition, and California’s leadership in the artificial intelligence sector.
(k) It is further the intent of the Legislature that this act complements and works alongside Assembly Bill 1405 of the 2025–26 Regular Session, and that the measures established by this act be interpreted in a manner that promotes consistency and coordination across California’s artificial intelligence governance framework.
SEC. 2.
Chapter 14 (commencing with Section 8898) is added to Division 1 of Title 2 of the Government Code, to read:CHAPTER 14. California Artificial Intelligence Standards and Safety Commission
8898.
As used in this chapter:(a) “Advanced safety standard” means a standard published by the agency or authorized by statute that establishes criteria for determining whether an AI system or model substantially exceeds minimum compliance requirements, incorporates emerging best practices, and achieves demonstrable outcomes that promote the safety, efficacy, reliability, and security of the system or model.
(b) “Agency” means the Business and Consumer Services Agency.
(c) “Artificial intelligence” or “AI” means an engineered or machine-based system that varies in its level of autonomy and that can, for explicit or implicit objectives, infer from the input it receives how to generate outputs that can influence physical or virtual environments.
(d) “AI auditor” means a person, partnership, academic institution, nonprofit, or corporation that conducts a covered audit on behalf of a third party.
(e) (1) Except as provided in paragraph (2), “commission”
means the California Artificial Intelligence Standards and Safety Commission established pursuant to Section 8898.1.
(2) If AB Assembly Bill 1709 of the 2025–26 Regular Session is enacted and establishes an e-Safety Advisory Commission, “commission” means the e-Safety Advisory Commission established pursuant to AB Assembly Bill 1709 of the 2025–26 Regular Session.
(f) “Covered audit” means either of the following:
an assessment of an AI system or model conducted in accordance with a minimum compliance standard or an advanced safety standard.
(1)An audit of an AI system or model conducted to verify whether the AI system or model complies with applicable state law.
(2)An assessment of an AI system or model conducted in accordance with an advanced safety standard.
(g) “Independent verification organization” or “IVO” means
an AI auditor that is designated by the agency as having demonstrated expertise in assessing whether AI systems or models substantially exceed minimum compliance requirements, incorporate emerging best practices, and achieve demonstrable outcomes that promote their safety, efficacy, reliability, and security.
(h) “Minimum compliance standard” means a standard published by the agency or created authorized by statute that establishes criteria for determining whether an AI system or model meets both of the
following meet minimum safety, efficacy, reliability, and security requirements necessary to comply with applicable state law. law:
(1) The operational effectiveness of an AI system or model.
(2) The design and implementation of internal controls established for an AI system or model.
8898.1.
(a) On or before July 1, 2027, the agency shall establish the California Artificial Intelligence Standards and Safety Commission that consists of all of the following members:(1) A representative, appointed by the Governor, from a business that develops or deploys AI systems or models and employs more than 100 workers.
(2) A representative, appointed by the Governor, from a business that develops or deploys AI systems or models and employs fewer than 100 workers.
(3) An expert from academia, appointed by the Governor, with experience
researching AI safety.
(4) An individual, appointed by the Governor, Speaker of the Assembly, representing the interests of civil society, including, but not limited to, nongovernmental organizations, public policy institutes, or consumer advocacy organizations.
(5) An individual, appointed by the Governor, Senate Rules Committee, representing the legal and economic interests of workers in relation to AI systems and models.
(6)The President of the California Board of Accountancy or that person’s designee.
(6) An individual authorized to practice public accountancy in this state, appointed by the Governor, who represents the accounting and auditing profession and has experience providing assurance services related to artificial intelligence and technology systems.
(7) The Attorney General or that person’s designee.
(b) Members appointed pursuant to this subdivision shall serve four-year terms and may be reappointed.
(c) If AB Assembly Bill 1709 of the 2025–26 Regular Session is enacted, and the e-Safety
Advisory Commission is established pursuant to that measure, this section shall be inoperative.
8898.2.
(a) The commission shall do all of the following:(1) Develop standards for the assessment of AI systems and models by AI auditors. In developing these standards, the commission shall do all of the following:
(A) Evaluate existing standards, frameworks, guidelines, criteria, and best practices developed or published by government agencies, standards-setting organizations, including national and international auditing and assurance
standard-setting organizations, professional accountancy bodies, AI auditors, private entities that develop or deploy AI systems or models, or other independent experts with relevant expertise.
(B) Establish the following two tiers of standards:
(i) Minimum compliance standards.
(ii) Advanced safety standards.
(C) Prioritize, at a minimum, developing standards for AI systems, models, and use cases that are subject to state law.
(D) Create a process for independent experts to submit advanced safety standards for evaluation by the commission pursuant to this chapter.
(2) Develop procedures for determining whether to suspend or terminate the registration of an AI auditor. In developing these procedures, the commission shall consider all of the following:
(A) Failures to adhere to appropriate standards, including standards developed pursuant to this chapter.
(B) Material misrepresentations in audit reports or required disclosures.
(C) Conflicts of interest that impair independence.
(D) Failure to maintain adequate documentation.
(E) Conduct that reasonably calls into question the integrity, objectivity,
or competence of the AI auditor.
(3) Develop criteria for determining whether an AI auditor qualifies as an independent verification organization. In developing these criteria, the commission shall consider whether the IVO is capable of all of the following with respect to an AI system or model in operation:
(A) Assessing the safety, efficacy, reliability, security, and robustness of the AI system or model.
(B) Identifying reasonably foreseeable risks arising from the development, deployment, operation, use, or misuse of the AI system or model.
(C) Verifying that the AI system or model has achieved and will maintain a level of risk that is appropriate in light of
the AI system’s or model’s intended purpose, foreseeable uses and misuses, affected populations, and the severity and likelihood of potential harms.
(D) Determining whether the AI system or model should be subject to ongoing monitoring, periodic reassessment, or other postdeployment review.
(E) Employing or otherwise engaging personnel with sufficient technical expertise to conduct covered audits.
(F) Identifying and managing potential conflicts of interest that may undermine the integrity, quality, or independence of covered audits.
(4) Publish any standards, procedures, and criteria
developed pursuant to this chapter in a publicly accessible format on the agency’s internet website. For each standard published pursuant to this paragraph, the commission shall identify, to the extent applicable, all of the following:
(A) Any state law to which the standard is intended to apply.
(B) Whether the standard is a minimum compliance standard or an advanced safety standard.
(C) The category, use case, or type of AI system or model to which the standard is intended to apply.
(D) The date upon which the standard was adopted and the date of the most recent revision to the standard.
(E) Any additional information necessary to facilitate the interpretation and application of the standard.
(5) Publish a statement on the agency’s internet website prominently disclosing that the publication of a standard pursuant to this subdivision does not constitute recommendation or endorsement by the state of any AI system or model.
(6) Review Regularly review standards, procedures, and criteria published pursuant to this chapter at least annually and revise them as appropriate to reflect changes in applicable state law, technological
developments, widely recognized industry standards, and emerging best practices.
(7) To the extent practicable, align any standards, procedures, and criteria developed pursuant to this section with existing professional and regulatory audit and assurance standards, frameworks, and criteria to promote alignment, consistency, comparability, and reliability across regulatory and industry frameworks.
(8) To the extent practicable, structure the requirements of this section to minimize duplicative compliance obligations, including by allowing reports, assessments, audits, or assurance engagements prepared to satisfy substantially similar requirements to be used for purposes of this section
if those reports, assessments, audits, or engagements satisfy the requirements of this section.
(b) In developing advanced safety standards pursuant to this chapter, the commission shall incorporate objective and clearly defined criteria for assessing all of the following with respect to an AI system or model in operation:
(1) The safety, efficacy, reliability, security, and robustness of the AI system or model.
(2) Reasonably foreseeable risks arising from the development, deployment, operation, use, or misuse of the AI system or model.
(3) The adequacy of any measures adopted to prevent, mitigate, and respond to those risks.
(4) Whether the AI system or model has achieved and maintained a level of risk that is appropriate in light of the AI system’s intended purpose, foreseeable uses and misuses, affected
populations, and the severity and likelihood of potential harms.
(5) Whether the AI system or model should be subject to ongoing monitoring, periodic reassessment, or other postdeployment review.
(6) Whether a developer, deployer, or other responsible entity has implemented and adhered to appropriate internal safety standards and protocols.
(7) Any other criteria specified in applicable state law.
8898.3.
(a) In carrying out its duties under this chapter, the commission(1) Certified public accountants and public accountants in good standing as certified by the California Board of Accountancy.
(2) AI
auditors that are not certified public accountants or public accountants.
(3) Independent verification organizations.
(4) Academic institutions.
(5) Private entities that develop or deploy AI systems or models.
(6) Consumer protection, labor, and civil society organizations.
(7) Relevant federal, state, and local agencies.
(8) National and international standards-setting organizations.
(9) Any other relevant stakeholders, as determined by the commission.
(b) The commission shall convene working groups to solicit stakeholder input in the development and revision of standards, procedures, and criteria pursuant to this chapter.
8898.4.
The commission may establish voluntary pilot programs to evaluate the effectiveness of standards developed pursuant to this chapter.8898.5.
(a) This chapter does not do either of the following:(a)
(1) Establish liability solely for failure to comply with a standard developed or published pursuant to this chapter.
(b)
(2) Constitute recommendation or endorsement by the state of any AI system or model.
(b) In an action alleging that a defendant’s development, modification, or use of an artificial intelligence system or model caused harm, the fact that an audit has been performed on the system or model in accordance with a standard published under this chapter is relevant to, but not conclusive of, the action.