An act to amend Section 22757.3.1 of the Business and Professions Code, relating to artificial intelligence.
AB 2713 amends California’s AI Transparency Act to broaden and clarify how large online platforms must handle system provenance data for content distributed on their services. The bill expands the law’s coverage from provenance data that is merely embedded in or attached to content to provenance data that is also otherwise associated with content. It also updates the required user-facing disclosures so platforms must indicate whether provenance data or digital signatures are embedded into, attached to, or otherwise associated with the content.
The bill preserves the existing framework that requires large online platforms to detect provenance data, provide a user interface disclosing its availability, and allow users to inspect it in an easily accessible manner. It restates the three permitted inspection methods: displaying provenance data in the platform’s interface, linking to a website or application that displays it, or allowing a download of the content with its provenance data, subject to federal copyright law and in a format that is not easily repurposed with unrelated content. It also extends the anti-stripping rule to cover provenance data or digital signatures associated with content downloaded from the platform, and it clarifies that the section does not require action for provenance data that does not comply with recognized standards. The operative date for these changes is January 1, 2027.
The bill’s practical impact is to impose more explicit provenance-detection and disclosure obligations on large online platforms operating in California, especially those hosting AI-generated or AI-altered media. It affects platform compliance systems, content labeling interfaces, and handling of standardized provenance metadata and digital signatures. Because the bill amends an existing statute rather than creating a new program, it primarily refines enforcement and implementation details for platforms subject to the California AI Transparency Act.
The general sentiment reflected in the available legislative history is favorable. The bill received a unanimous 15-0 do-pass-as-amended vote in committee and advanced to third reading, suggesting broad support for strengthening AI content transparency and provenance protections. No committee transcript is provided, but the vote pattern indicates little visible opposition at this stage.
The main points of contention appear to be technical and implementation-related rather than ideological. The bill’s language raises questions about how platforms will detect provenance data that is “associated with” content, how they will present that information to users, and how the download option interacts with federal copyright law. Another likely issue is the burden on large online platforms to maintain provenance integrity without stripping metadata, particularly where content is downloaded, reposted, or transformed. The bill also leaves open that nonstandard provenance systems are outside its requirements, which may matter to platform operators and standards bodies.
AB 2713 would amend Section 22757.3.1 of the Business and Professions Code, expanding and clarifying the obligations imposed by the California AI Transparency Act on large online platforms. It broadens the statute’s provenance-data language to include content that is embedded into, attached to, or otherwise associated with provenance data or digital signatures, and it updates the required user disclosures and inspection methods accordingly. The bill also extends the prohibition on knowingly stripping provenance data or digital signatures to content downloaded from the platform, while preserving the statute’s focus on widely adopted, standards-based provenance systems. The operative date for the amended requirements would be January 1, 2027, affecting platform compliance practices, content labeling, and metadata handling for AI-generated and AI-modified media distributed in California.
The available legislative history suggests generally positive sentiment toward the bill. It passed the relevant committee on a 15-0 do-pass-as-amended vote, indicating unanimous support among voting members at that stage. The absence of recorded opposition or committee transcript discussion suggests the measure was viewed as a technical refinement to existing AI transparency law rather than a controversial policy shift.
Any disagreement appears centered on implementation details rather than the bill’s core goal of AI transparency. The most notable issues are the expanded definition of provenance data as embedded in, attached to, or otherwise associated with content, the requirement that platforms detect and disclose that information, and the copyright-sensitive download option. Large online platforms may be concerned about compliance costs, technical feasibility, and the risk of overbroad obligations, while supporters are likely focused on ensuring users can identify AI-generated or altered content and preserve provenance metadata. The bill also limits its requirements to provenance data that conforms to widely adopted standards, which may be a point of discussion for stakeholders working on nonstandard systems.