AI Frontier Institute

Independent verification organizations.

View primary source →

California Senate Bill 813 (Chapter 179), signed September 9, 2026, adds Chapter 14 (Sections 8898–8898.4) to the Government Code to establish a framework for 'independent verification organizations' (IVOs): AI auditors designated by the Government Operations Agency as having demonstrated expertise in assessing the risks posed by an AI system or model and identifying the metrics and methodologies behind that assessment. An 'AI auditor' may be a person, partnership, academic institution, nonprofit, or corporation (Section 8898(c)). By January 1, 2028, the Agency must develop application requirements for designation (applicant qualifications, proposed benchmarks and methodologies, supporting documentation), procedures for suspending or terminating a designation (including for conflicts of interest that impair independence, material misrepresentations, and cybersecurity lapses), and designation criteria, all published on its website (Section 8898.1). The criteria must at minimum address risk-assessment capability, technical personnel expertise, conflict-of-interest management, and independence from the party being assessed and its affiliates, meaning no operational or management dependence and freedom from the assessed party's control over conclusions. An IVO may accept payment from a party it assesses at reasonable market rates but may not accept terms that condition payment on the results (Section 8898.1(c)(2)(C)-(D)). The Agency must consult broadly and convene working groups that include engineers from competing AI companies and AI safety experts, and report the working groups' findings to the Legislature (Section 8898.2). Designated IVOs must report annually to the Agency and Legislature on their standards and methodologies and on any changes to governance or funding sources relevant to independence, with redactions permitted for trade secrets and security and unredacted records retained for five years (Section 8898.3). The chapter creates no liability for failing to meet a standard, is not a state endorsement of any AI system, does not require any developer or deployer to engage an IVO, and does not require an IVO to conduct compliance audits to register; an audit performed under an identified standard is relevant but not conclusive in litigation alleging AI-caused harm (Section 8898.4). Executive Order N-9-26 (September 18, 2026) directs the Agency to complete the Section 8898.1 requirements by May 1, 2027, eight months ahead of the statutory deadline.

← Back to the tracker