When does California Generative AI: Training Data Transparency take effect?
California Generative AI: Training Data Transparency has been effective since January 1, 2026.
Compliance reference — obligations, penalties, applicability, and primary sources.
Last verified July 27, 2026
California AB 2013 — the Generative AI Training Data Transparency Act, also searched as "CA AB 2013" or "AB 2013 California" — is codified at Civil Code Title 15.2, Sections 3110 and 3111. It requires a developer of a generative AI (GenAI) system or service that was released on or after January 1, 2022 and made publicly available to Californians to publish training-data documentation on the developer's public website. AB 2013 regulates developers, not deployers or end users: the disclosure duty falls on the entity that designs, codes, produces, or substantially modifies the GenAI system or service.
AB 2013's training-data documentation duty is one part of California's generative-AI transparency regime; it operates alongside California SB 942, the AI Transparency Act, which governs disclosure of AI-generated image, video, and audio content, and California SB 53 (TFAIA), which imposes safety-framework and critical-incident-reporting duties on frontier-model developers.
The statute reaches a developer that makes a covered GenAI system or service available to people in California, regardless of where the developer is located. A "developer" includes a person that substantially modifies a GenAI system or service, so fine-tuning or materially retraining a third-party model can bring an organization into scope.
The developer must post a high-level summary of the datasets used to develop the GenAI system or service. The summary must address each of the following:
The first posting deadline is January 1, 2026. Because the requirement is keyed to public availability, a developer that releases a new covered model — or substantially modifies an existing one — after that date must post updated documentation before that release or modification reaches Californians. The obligation is therefore continuous, not a one-time filing.
Civil Code § 3110 excludes a GenAI system or service whose sole purpose is to help ensure security and integrity, the operation of aircraft in the national airspace, or that is developed for national security, military, or defense purposes and is made available only to a federal entity.
Civil Code §§ 3110–3111 do not state a standalone civil-penalty amount or create an express private right of action. The Atlas therefore records no fixed statutory penalty for AB 2013 rather than importing a dollar amount or enforcement theory from another California law.
Primary sources: California Civil Code Title 15.2, §§ 3110–3111, leginfo.legislature.ca.gov, and the chaptered AB 2013 text — retrieved July 27, 2026.
Specific compliance requirements derived from the primary source. Each item links to the relevant statutory section where applicable.
Publicly post on the developer's website a high-level summary of training datasets used for any generative AI system or service made available to Californians on or after January 1, 2022.
Deadlineby_2026-01-01
Address each statutory element in the dataset summary — sources or owners, how the data furthers the system's purpose, number and types of data points, copyright/trademark/patent or public-domain status, purchased-or-licensed status, presence of personal information or aggregate consumer information, any cleaning or modification, first-use dates, collection time period, and any use of synthetic data generation.
Deadlineby_2026-01-01
Refresh and re-post the public training-data summary before each subsequent public release or substantial modification of a covered generative AI system, so the disclosure stays current rather than ending at the initial January 1, 2026 deadline.
Deadlinebefore_each_release
Use the side-by-side comparisons to see how this law interacts with adjacent US AI laws and governance frameworks.
Compare source-backed California AI law records and compliance categories.
A single law is one input to a broader program. These guides show how to assign accountability, map every obligation to a control, and split developer-versus-deployer duties across the systems a team operates.
The operating model — owners, committee, decision rights, lifecycle gates — that turns a law into an accountable, auditable program anchored to the NIST AI RMF GOVERN function.
How to map each AI system to the laws it triggers, the controls that satisfy them, and the evidence that proves the controls operate.
Which duties attach to building an AI system versus putting one into use — the split most state AI laws turn on.
California Generative AI: Training Data Transparency has been effective since January 1, 2026.
No fixed cap is specified; remedies follow the underlying enforcement framework.
California Generative AI: Training Data Transparency applies to developer of AI systems within its jurisdictional scope.
Every fact above is sourced from the official primary source. Independent verification recommended before acting on the information.
Last reviewed July 27, 2026. Reviewed by the AI Compliance Atlas editorial process against primary sources. Source selection, retrieval dates, and update rules are documented in the Atlas methodology.
We may receive referral commissions from recommended compliance tools. Recommendations are based on product fit and not on commission size. Links marked “partner link” include a tracked redirect.