California Senate Bill 813, authored by Sen. Jerry McNerney, was approved by Governor Gavin Newsom and filed with the Secretary of State on 9 September 2026, chaptered as Chapter 179 of 2026. It adds Chapter 14 to Division 1 of Title 2 of the Government Code, Sections 8898 through 8898.4, under the heading "California Artificial Intelligence Safety Independent Verification Organizations." The chaptered description is "Independent verification organizations." The enacted law directs the Government Operations Agency to build a designation regime for AI auditors, on or before 1 January 2028, and leaves participation voluntary. For the bill's legislative history, sponsorship and reception, see California SB 813 (AI independent verification organizations).
What the enacted text does not contain
The enacted law differs from the version the bill's public record described through mid-2026 in three respects that change its structure.
There is no commission. The July 2, 2026 amended version would have established a California Artificial Intelligence Standards and Safety Commission, to be constituted by the Business and Consumer Services Agency on or before July 1, 2027, and would have assigned the substantive duties to that commission. The enacted text assigns them directly to the Government Operations Agency, defined at Section 8898(a) as "the agency," and sets the deadline at 1 January 2028.
The AB 1709 contingency — under which the duties would pass to the e-Safety Advisory Commission if that measure were enacted — does not appear.
No contingency on AB 1405 appears either. The sponsors had stated that both measures must be enacted for the framework to operate, and the July 2 version made SB 813's operation expressly conditional on AB 1405's enactment. Chapter 179 contains no such condition. Whether it was removed or relocated into AB 1405 cannot be determined from this text; AB 1405's enacted text is not held.
Definitions
| Term | Section | Definition |
|---|---|---|
| Agency | 8898(a) | The Government Operations Agency |
| Artificial intelligence | 8898(b) | "[A]n 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" |
| AI auditor | 8898(c) | "[A] person, partnership, academic institution, nonprofit, or corporation that conducts a covered audit on behalf of a third party" |
| Covered AI audit | 8898(d) | "[A]n audit conducted to assess internal controls, processes, or systems implemented for an AI system or model that are necessary for compliance with state law" |
| Independent verification organization (IVO) | 8898(e) | An AI auditor "designated by the agency as having demonstrated expertise in assessing the risks posed by an AI system or model and identifying the metrics and methodologies that form the basis for that assessment" |
The agency's duties (Section 8898.1)
On or before 1 January 2028 the Government Operations Agency must do all of the following.
Application requirements (subdivision (a)), under which an applicant must submit its qualifications demonstrating competence; information sufficient to address the published designation criteria; "[t]he benchmarks, technologies, metrics, and methodologies the IVO proposes using to conduct the IVO's work"; and any documentation needed for the agency to verify the application's accuracy.
Suspension and termination procedures (subdivision (b)), which must consider six enumerated grounds: failures to adhere to appropriate standards; material misrepresentations in the application, audit reports or required disclosures; conflicts of interest that impair independence; failure to maintain adequate documentation; "[c]onduct that reasonably calls into question the integrity, objectivity, or competence of the IVO"; and lapses in cybersecurity.
Qualification criteria (subdivision (c)), in developing which the agency must identify and consider existing standards, frameworks, guidelines and best practices from government agencies, standards-setting organizations including national and international auditing and assurance organizations, AI auditors, AI developers and deployers, and other independent experts. At a minimum the agency must consider whether the IVO meets four criteria with respect to a system in operation: assessing its risks and identifying the metrics and methodologies forming the basis of that assessment; employing or engaging personnel with sufficient technical expertise; identifying and managing conflicts of interest; and maintaining independence from the assessed party.
Publication, disclaimer, review and alignment (subdivisions (d) through (h)): publishing the requirements and criteria in a publicly accessible format; publishing a statement "prominently disclosing that the publication of requirements or criteria pursuant to this subdivision does not constitute recommendation or endorsement by the state of any AI system or model"; reviewing and revising the requirements to reflect changes in state law, technological developments, recognised industry standards and emerging best practices; aligning the procedures and criteria with existing professional and regulatory audit and assurance standards "to promote consistency, comparability, and reliability"; and structuring the requirements "to minimize duplicative compliance obligations," including by allowing reports, assessments, audits or assurance engagements prepared for substantially similar requirements to be used for this chapter's purposes.
Independence and the contingent-fee prohibition
The two independence criteria at Section 8898.1(c)(2) are the operative answer to the developer-pays objection, and they permit the arrangement while removing its sharpest edge.
Subdivision (C) requires the agency to consider whether the IVO identifies and manages conflicts of interest that may undermine its integrity, quality or independence, "including ensuring financial relationships do not impact the judgment and decisionmaking of the IVO." It then states the rule directly: "An IVO may accept payment from a party being assessed at reasonable market rates but shall not accept terms which condition any payment or the amount of any payment on the results of their assessment."
Subdivision (D) requires independence from the assessed party "including by having no operational or management dependence on the party being assessed or its affiliates and by remaining otherwise free from the assessed party's control in reaching conclusions or making recommendations, including, as appropriate, through contractual safeguards and conflict-of-interest policies."
Payment by the assessed developer is therefore permitted. What is prohibited is outcome-contingent payment. That narrows the conflict described on Independent Verification Organizations (IVOs) without eliminating the structural feature it rests on — the auditor's repeat-business relationship with the party it assesses.
Consultation and reporting
Section 8898.2(a) requires the agency to consult as appropriate with AI auditors; independent verification organizations; academic institutions; private entities that develop AI systems or models, "including startups and other emerging technology companies"; private entities that deploy or operate them; "[c]onsumer protection, labor, and civil society organizations"; relevant federal, state and local agencies; national and international standards-setting organizations; and any other stakeholders the agency determines.
Section 8898.2(b)(1) requires the agency to convene working groups to solicit stakeholder input on the identification of standards and the development and revision of procedures, requirements and criteria. Those groups "shall include, but not be limited to, engineers from AI companies that are competitors and AI safety experts." Section 8898.2(b)(2) requires a report to the Legislature under Section 9795 on the working groups' findings.
Section 8898.3(a) requires each designated IVO to report annually, and no sooner than 12 months after initial designation, to both the agency and the Legislature, covering summaries of its standards and methodologies; a description of any changes to its governance policies or funding sources relevant to its conflicts of interest or independence; and any changes to its application information. Subdivision (b) permits redactions necessary to protect trade secrets, cybersecurity, public safety or US national security, or to comply with federal or state law. Subdivision (c) requires the IVO to "describe the character and justification of the redaction in any published version of the document to the extent permitted by the concerns that justify redaction" and to retain the unredacted information for five years.
Limits and the evidentiary rule (Section 8898.4)
Subdivision (a) states four things the chapter does not do:
- Establish liability solely for failure to comply with a standard under the chapter.
- Constitute recommendation or endorsement by the state of any AI system or model.
- Require any developer, deployer or operator "to engage an IVO or to undergo a covered AI audit as a condition of developing, deploying, or operating an AI system or model in this state."
- Require an IVO to conduct compliance audits in order to register with the agency.
Subdivision (b) is the chapter's only provision touching liability: "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 in accordance with a standard identified under this chapter is relevant to, but not conclusive of, the action."
That is an evidentiary rule rather than a safe harbour. It admits the audit as relevant and denies it conclusive effect, and on its face favours neither party: a conforming audit is evidence a defendant may adduce, and the absence of one is not made a basis of liability by subdivision (a)(1). An account describing the law as conferring a liability shield on audited developers, or as making the audit legally inert, would misstate it in opposite directions. It is this provision that the certification-for-shield critique at Don't Let AI Developers Hire Their Own Referees (Weil, July 2026) turns on.
Digest Key
Majority vote; no appropriation; fiscal committee yes; no local program.
Provenance
Captured on 2026-09-12 from leginfo.legislature.ca.gov, complete: Legislative Counsel's Digest, Digest Key, and the full operative text of Sections 8898 through 8898.4. Thirteen divergences and cautions against the secondary accounts the vault previously relied on are recorded in the raw file's ## Source-fidelity notes; the structural ones are the absence of the commission, the change of administering agency and deadline, and the absence of the AB 1709 and AB 1405 contingencies.
Relationships
- supersedes: California SB 813 (AI independent verification organizations) — as to the bill's operative content, which the page described from the July 2, 2026 amended version
- related: California AB 1405 (AI auditor registry) — the paired auditor registry, whose enacted text is not held and whose relationship to this chapter the enacted text does not state
- depends-on: Independent Verification Organizations (IVOs) — the auditor model the chapter designates
- contradicts: Don't Let AI Developers Hire Their Own Referees (Weil, July 2026) — the enacted text permits developer payment while prohibiting outcome-contingent payment, and creates no shield
- related: AI Liability — the relevant-but-not-conclusive evidentiary rule at Section 8898.4(b)
- related: State-Level AI Regulation — a California entry in the state AI-legislation landscape
- related: California SB 53 — the Transparency in Frontier Artificial Intelligence Act, cited in the Legislative Counsel's Digest as existing law
Sources
- Raw source:
Raw Sources/California SB 813 - enrolled text as signed (Chapter 179, 2026-09-09).md - Enrolled text: leginfo.legislature.ca.gov