 |
ARIZONA STATE SENATE
Fifty-Seventh Legislature, Second Regular Session
AMENDED
artificial intelligence; content verification
Purpose
Effective February 1, 2027, requires certain providers of generative artificial intelligence (AI) systems to include provenance data in specified media that is created or materially altered by the system.
Background
The federal National Artificial Intelligence Initiative Act of 2020 (Act) codified the establishment of a national AI initiative and associated federal offices and committees. The Act directed the Secretary of Commerce, in coordination with other federal agencies, including the National Institute of Standards and Technology, the Department of Energy and the Department of Homeland Security, to establish guidelines and best practices for developing safe, secure and trustworthy AI systems, with the aim of promoting consensus industry standards (15 U.S.C. § 9401 et seq).
Executive Order 14179 deems that it is the policy of the United States to sustain and enhance AI to promote economic competitiveness and national security. The Order directed various federal agencies and individuals to develop an AI action plan and identify actions taken pursuant to Executive Order 14110 that conflict with this policy (Exec. Order No. 14179). Executive Order 14365 deems that it is the policy of the United States to achieve a minimally burdensome national policy framework for AI. The Order directed the establishment of an AI Litigation Task Force for the purpose of evaluating and challenging state laws inconsistent with this policy and restricted funding to the states under the Broadband Equity Access and Development Program (Exec. Order 14365).
There is no anticipated fiscal impact to the state General Fund associated with this legislation.
Provisions
1.
Requires a covered provider to include provenance data in any video,
image or audio content that is created or materially altered by the covered
provider's generative AI system and that is available to be shared with users
outside of the system.
2. Specifies that this requirement does not apply to text content that is generated as a result of an interactive experience, including voice assistant responses, virtual reality environments, augmented reality applications, real-time conversational AI interactions and other real-time, dynamic interactions between a user and a generative AI system.
3. Defines covered provider as a person or entity that creates, codes or otherwise produces a generative AI system that has more than one million monthly users and that is publicly accessible within the geographic boundaries of Arizona for personal use by consumers.
4. Specifies that the definition for covered provider does not include:
a) cloud service providers;
b) providers that distribute, or make available, a generative AI system that is primarily licensed or sold for business to business purposes;
c) products, services or applications that are primarily marketed and intended to facilitate accessibility for a person with a disability if the developer takes reasonable measures to protect against unauthorized use;
d) internet service providers;
e) internet service providers' affiliates or subsidiaries;
f) internet search engines.
5. Requires a covered provider to use commercially and technically reasonable methods to make the provenance data difficult to remove or tamper with.
6. Stipulates that the above requirement may be satisfied with the use of a commonly supported technical standard for watermarking and metadata.
7. Stipulates that content is not considered materially altered if the alteration consists of a minor modification that does not lead to significant changes to the perceived content or meaning of the content.
8. Prohibits covered providers from including any information that relates to an identified individual, or a reasonably identifiable individual, in provenance data that is in any video, image or audio content created or materially altered by the covered provider's generative AI system, to the extent that it is commercially and technically reasonable, unless a covered provider chooses to include the information where a user affirmatively elects to include the information.
9. Stipulates that requirements relating to content verification for generative AI systems do not:
a) apply to any product, service, internet website or application that either:
i. exclusively provides video games or non-user-generated television, music, streaming, movie or interactive experiences; or
ii. provides for the sale of goods or services directly to consumers through the internet and allows customers to browse, select and purchase items virtually.
b) require a covered provider to disclose any trade secret or confidential or proprietary information about the design or use of an AI system.
10. Defines generative AI system as an AI system that can generate synthetic content.
11. Defines minor modification to include changing brightness, contrast or color, sharpening, saturating, applying filters, resizing, scaling, cropping, format conversions, resampling, denoising and removing background noise in audio.
12. Defines provenance data as data that is embedded into digital content or that is included in the digital content's metadata to verify the digital content's authenticity, origin or history of modification and that includes generally accepted security best practices.
13. Contains a statement of legislative findings.
14. Becomes effective on February 1, 2027.
Amendments Adopted by the House of Representatives
1. Limits the applicability of requirements for generative AI system provenance data to content that is available to be shared outside of the system.
2. Exempts various interactive experiences with generative AI systems from provenance data requirements.
3. Exempts various products and services from generative AI content verification requirements in general.
4. Excludes specified entities and services from the definition of a covered provider of generative AI systems.
5. Modifies the definition of covered provider to only include generative AI systems that have more than one million monthly users.
6. Prohibits, rather than requires, generative AI systems from including information on identifiable individuals in provenance data, but only to the extent that it is commercially and technically reasonable, unless a covered provider chooses to include the information where a user affirmatively elects to include the information.
7. Adds a delayed effective date of February 1, 2027.
8. Modifies the definitions of generative AI system and provenance data.
9. Makes conforming changes.
Senate Action House Action
JUDE 2/18/26 DP 4-3-0 AII 3/26/26 DPA 4-2-0-1
3rd Read 3/3/26 17-13-0 3rd Read 4/15/26 33-20-7
Prepared by Senate Research
April 15, 2026
ZD/ci