California AI Audit Laws: What SB 813 and AB 1405 Mean for AI Governance

California’s new AI audit laws do not require every organization using AI to obtain an independent audit. They do, however, establish a more formal accountability framework for the auditors that assess AI systems for compliance with state law.

Signed on September 9, 2026, SB 813 requires California to develop standards for designating independent verification organizations by January 1, 2028. AB 1405 requires the state to establish an AI Auditor Registry and, beginning January 1, 2029, prohibits unregistered parties from conducting covered AI audits. The law also introduces requirements related to auditor independence, methodology, documentation, evidence, and professional judgment.

For privacy, legal, compliance, security, and AI governance leaders, the immediate lesson is not that every AI system must now be audited. It is that independent scrutiny depends on reliable internal evidence. Organizations need to know which AI systems they use, which requirements apply, who approved key decisions, how controls operate, and how incidents and exceptions are resolved.

California’s framework signals a practical shift from stating AI principles to demonstrating how those principles are applied.

What California’s AI Audit Laws Do and Do Not Require

SB 813 and AB 1405 regulate different parts of California’s emerging AI assurance ecosystem.

SB 813 directs the California Government Operations Agency to establish criteria for designating independent verification organizations. AB 1405 creates a registry and operating requirements for auditors performing covered AI audits related to compliance with California law.

Neither law on its own requires every organization developing, deploying, or operating AI to undergo an audit. The direct statutory obligations fall primarily on California agencies, AI auditors, and designated independent verification organizations.

For organizations using AI, the significance is indirect but practical: when an audit is required under another law or undertaken voluntarily, the review will depend on documented controls, reliable evidence, clear ownership, and decisions that can be independently evaluated.

How SB 813 Establishes Independent AI Verification

SB 813 directs the Government Operations Agency to create a framework for designating independent verification organizations, or IVOs.

An IVO is 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 supporting that assessment.

By January 1, 2028, the agency must develop:

• Application requirements for organizations seeking IVO designation
• Qualification criteria addressing technical expertise and assessment capabilities
• Procedures for suspending or terminating an IVO’s designation
• Standards for identifying and managing conflicts of interest
• Requirements intended to preserve independence from the organization being assessed

The agency must convene stakeholder working groups and periodically update its requirements and criteria to reflect changes in state law, technology, recognized standards, and emerging practices.

Designated IVOs will also be required to submit annual reports describing their standards and methodologies, relevant changes to governance or funding, and updates to their application information.

Importantly, SB 813 expressly states that it does not require every AI developer, deployer, or operator to engage an IVO or undergo a covered AI audit as a condition of operating in California. The law establishes infrastructure for qualified independent verification; it does not create a universal audit mandate.

How AB 1405 Regulates Covered AI Audits

AB 1405 creates a complementary registration and accountability framework for AI auditors.

By January 1, 2029, the Government Operations Agency must establish a public AI Auditor Registry. Beginning on that date, a person or organization may not offer, sell, or conduct a covered AI audit unless registered with the agency.

The law defines a covered AI audit as an assessment of the internal controls, processes, or systems implemented for an AI system or model that are necessary for compliance with California law.

Registered auditors must disclose information that includes:

• The California laws and regulations under which they conduct covered audits
• Relevant certifications and accreditations
• The standards they apply
• Their standard operating procedures
• The basis for claims about the accuracy, reliability, or validity of their protocols

A registered auditor conducting a covered AI audit must provide the audited organization with a report that addresses the engagement’s scope and objectives, results, identified deficiencies, potential corrective measures, material limitations, and gaps in the available evidence.

The law also establishes requirements related to independence, objectivity, competence, record retention, and whistleblower protection. Registered auditors generally must retain covered audit reports and supporting documentation for at least 10 years.

These requirements move AI auditing closer to mature assurance disciplines built on defined scope, consistent methodology, documented evidence, disclosed limitations, and accountable professional judgment.

Who Is Directly Affected?

The two laws primarily regulate:

• AI auditors conducting covered audits related to California law
• Organizations seeking designation as IVOs
• California agencies responsible for establishing and administering the programs

Organizations that develop, deploy, or use AI are not automatically required to obtain an audit under SB 813 or AB 1405.

However, those organizations may still be affected when an audit is required under another California law, requested by a customer or business partner, included in a contractual obligation, or pursued voluntarily to strengthen assurance.

In each case, the organization’s ability to support an audit will depend on the quality of its internal governance records.

Why Organizations Should Prepare Before an Audit Is Required

An independent auditor can only evaluate evidence that an organization can identify, retrieve, and explain.

Organizations should be prepared to show:

• Which AI systems, models, vendors, and use cases are in scope
• Who owns each system and its associated risks
• What personal, confidential, or regulated data the system processes
• Which laws, policies, contracts, and control frameworks apply
• What testing, monitoring, and human oversight controls are operating
• How exceptions, incidents, failures, and remediation decisions are recorded
• Why deployment or continued use was approved
• Whether another reviewer could follow the evidence and understand the conclusion

An AI policy may establish expectations, but it does not prove those expectations are consistently applied. Audit readiness requires an operating model that connects inventory, assessment, controls, decisions, monitoring, incidents, and remediation.

Five Steps to Improve AI Audit Readiness

Organizations do not need to wait for the state’s IVO framework or auditor registry to begin strengthening their governance evidence.

1. Maintain a Complete AI Inventory

Create a centralized record of AI systems, models, vendors, embedded AI features, use cases, owners, affected individuals, and data flows.

The inventory should include AI capabilities built into third-party products, not only systems developed internally. It should also be updated as vendors, models, data sources, and uses change.

2. Apply Consistent Risk-Classification Criteria

Assess AI systems against applicable regulations, internal policies, contractual requirements, and control frameworks.

Comparable use cases should be evaluated using consistent logic. When two systems receive different classifications or controls, the organization should be able to explain why.

3. Preserve Decision Rationale

An approval alone provides limited evidence of diligence.

Document the facts considered, the evidence reviewed, the controls required, the exceptions approved, the stakeholders involved, the known limitations, and the reasons for the final decision. This creates a record that internal leaders and independent reviewers can follow.

4. Connect AI Incidents to Governance

Unexpected model behavior, harmful outputs, unauthorized data use, security failures, and agent activity outside an approved scope may reveal weaknesses in an organization’s controls.

AI incident response should feed lessons back into system assessments, monitoring plans, approval conditions, policies, and remediation decisions. That connection helps ensure governance evolves as actual risks emerge.

5. Establish Criteria for Selecting an AI Auditor

Organizations engaging an AI auditor should evaluate the auditor’s qualifications, technical expertise, independence, methodology, conflicts of interest, evidence requirements, and approach to documenting limitations.

California’s framework makes clear that an audit label alone is not enough. The credibility of an audit depends on the auditor’s independence and the rigor of the underlying process.

Independent Assurance Does Not Replace Internal Accountability

Third-party review can strengthen trust, but it cannot replace governance within the organization.

An audit examines a defined scope during a defined period. AI systems continue to change after that review. Models are updated, vendors modify their services, data sources evolve, employees discover new uses, and controls may become less effective.

Organizations therefore need continuous ownership, monitoring, incident response, and reassessment—not simply a favorable audit report.

Independent verification should be one component of a broader operating model that includes inventory, risk classification, accountable approvals, ongoing monitoring, incident management, remediation, and documented reassessment.

How RadarFirst Supports Defensible AI Governance

California’s new laws reinforce a central governance principle: organizations need evidence that their AI requirements are being applied in practice.

Radar AI Risk helps teams maintain a centralized AI inventory, evaluate systems against relevant policies and regulatory frameworks, assign accountable owners, and preserve the evidence and rationale behind governance decisions.

By connecting assessment, approval, monitoring, incident response, and reassessment through structured workflows, RadarFirst helps privacy, legal, compliance, risk, security, and business teams work from the same facts. The result is faster, more consistent decision-making and a clearer record of diligence when internal or independent reviewers ask how an AI system is governed.

California may be establishing the first state framework of its kind, but the underlying expectation is familiar to regulated organizations: if an organization says an AI system is governed, safe, or compliant, it should be prepared to show the evidence supporting that conclusion.

Request a demo to see how RadarFirst can help your teams maintain an up-to-date AI inventory, apply requirements consistently, link incidents to governance, and produce audit-ready decision records.

This article is provided for informational purposes and does not constitute legal advice.