The Generative Artificial Intelligence Accountability Act (SB 896, 2023–2024 Regular Session) is a California statute governing the use of generative AI by state agencies and departments. It requires disclosure when generative AI is used to communicate with the public about government services, an annual critical-infrastructure risk analysis by the state's emergency-management agency, training for state employees, and use of a state-approved set of generative AI tools. Authored by Sen. Bill Dodd (D-Napa) and signed by Governor Gavin Newsom on September 29, 2024, it functions as the California-state-government analog of the federal OMB Memorandum M-24-10.
Status and timeline
SB 896 was enacted in the California Legislature's 2023–2024 Regular Session, authored by Sen. Bill Dodd (D-Napa) and signed by Governor Newsom on September 29, 2024. It took effect on January 1, 2025. The act continues work begun under Newsom's Executive Order N-12-23 (September 6, 2023) on generative AI, which it treats as a companion instrument. The annual critical-infrastructure risk reporting to the Legislature is an ongoing obligation.
The act formed part of California's 2024 generative-AI legislative package alongside California AI Transparency Act (SB 942) (provider-side), California AB 3030 (Healthcare AI Disclosure) (healthcare), California SB 243 — Companion Chatbots (minors and companion chatbots), and California SB 53 — Transparency in Frontier AI Act (frontier developers).
Scope
The act applies to California state agencies and departments using generative AI, mirroring at the state level the federal architecture of OMB Memorandum M-24-10. It does not directly regulate private-sector AI, but shapes the state's AI procurement and workforce posture. It does not address municipal or local government use; cities and counties are not covered.
Key provisions
Citizen-facing disclosure
When a state agency or department uses generative AI to directly communicate with a person regarding government services and benefits, the communication must include a disclaimer indicating that it was generated by generative AI, together with information describing how the person may contact a human employee of the agency or department. The requirement parallels California AB 3030 (Healthcare AI Disclosure)'s healthcare-provider structure, applied to state government.
Critical-infrastructure risk analysis
The California Office of Emergency Services (OES) must perform an annual risk analysis of potential threats posed by the use of generative AI to California's critical infrastructure, including threats that could lead to mass casualty events. OES must provide a high-level summary of the analysis annually to the Legislature, and must coordinate with owners and operators of critical infrastructure and with relevant federal agencies. This duty runs alongside California SB 53 — Transparency in Frontier AI Act's critical-safety-incident reporting to OES, placing OES at the center of the state's AI critical-infrastructure and catastrophic-risk monitoring.
State-employee training
Agencies must make available trainings for state-worker use of state-approved generative AI tools. The training must address equitable outcomes, output inaccuracies, fabrications or hallucinations, privacy risks, and biases. It is developed in collaboration with the Government Operations Agency (CalGovOps), the Department of Technology (CDT), the California Privacy Protection Agency (CPPA), and state-employee representative organizations.
State-approved tool program
Agencies must use state-approved generative AI tools, with procurement channels centralized to reduce shadow-IT risk. The program continues the work begun under Executive Order N-12-23 (September 6, 2023).
Workforce and education
The act encourages state investment in AI education and the building of AI competency in the state workforce, and directs state training providers to develop curricula for generative-AI literacy.
Annual legislative reporting
CalGovOps, CDT, and CPPA must report annually on state generative-AI use, including agency-level adoption, training, incidents, and public-facing use cases.
Relation to other instruments
SB 896 is narrower in scope than OMB Memorandum M-24-10 — it has no rights- or safety-impacting taxonomy and no minimum-practices floor — but it institutionalizes disclosure and workforce practices and creates a critical-infrastructure feedback loop to OES. The table below situates it among related instruments tracked in adjacent pages.
| Instrument | Applies to | Central structure |
|---|---|---|
| CA SB 896 | California state government | Disclosure + training + OES risk analysis + approved-tool program |
| OMB Memorandum M-24-10 | Federal civilian agencies | CAIO + minimum practices + AI use-case inventory |
| OMB Memorandum M-24-18 | Federal AI acquisition | Vendor / contract requirements |
| Utah Artificial Intelligence Policy Act (SB 149) | Utah consumer interactions generally | Disclosure + liability + sandbox |
| California AB 3030 (Healthcare AI Disclosure) | Healthcare providers using GenAI | Patient-communication disclaimer |
| California SB 53 — Transparency in Frontier AI Act | Frontier model developers | Safety framework + transparency + incident reporting to OES |
A side-by-side comparison with OMB Memorandum M-24-10, the federal-executive-branch instrument it most closely tracks:
| Dimension | CA SB 896 | OMB M-24-10 |
|---|---|---|
| Level | State executive | Federal executive |
| Core duty | Citizen-facing disclosure + training + approved tools | CAIO + AI governance board + inventory + minimum practices |
| Rights-impacting framework | Not explicitly | Yes (presumed categories + mitigations) |
| Safety-impacting framework | Not explicitly; OES risk analysis fills adjacent role | Yes (minimum practices) |
| Waiver mechanism | Implicit via approved-tool program | Explicit CAIO waiver |
| Inventory / transparency | Annual reports | Annual AI use-case inventory (public) |
M-24-10 is the more structured and prescriptive of the two; SB 896 is lighter-touch but adds a mass-casualty and critical-infrastructure safety dimension that the federal instrument does not contain.
Secondary commentary has described SB 896 as the first US state-government generative-AI governance statute codified into law, predating most state-level executive orders on state-agency AI, and has characterized its critical-infrastructure mass-casualty risk analysis as a new role for a state emergency-management agency in AI catastrophic-risk monitoring. Other states have followed with variations, including Washington HB 1170, Minnesota HF 1146, and New Jersey A 3854.
Relationships
- related: OMB Memorandum M-24-10 — federal-executive-branch analog
- related: OMB Memorandum M-24-18 — federal AI-acquisition analog
- related: California SB 53 — Transparency in Frontier AI Act — shared OES critical-safety-incident infrastructure
- related: California AI Transparency Act (SB 942), California AB 3030 (Healthcare AI Disclosure), California SB 243 — Companion Chatbots — 2024 CA GenAI package
- related: Utah Artificial Intelligence Policy Act (SB 149) — first state to institute GenAI disclosure + AI Office of Policy
- related: US AI Regulatory Approaches Compared — adds state-executive-branch dimension
Sources
- Bill text: https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240SB896
- Secondary: CalMatters Digital Democracy SB 896 page; StateScoop (2024); California Air Resources Board chaptered-bill summary; Independent Voice (Oct 2, 2024); A&O Shearman "Zooming in on AI #9"; ETO AGORA instrument entry
- Related: Executive Order N-12-23 (Newsom, Sep 6, 2023)