Artificial Intelligence. Artificial intelligence (AI) is defined and understood in different ways. The National Institute of Standards and Technology (NIST), in its AI Risk Management Framework, defines an AI system as an engineered or machine-based system that can, for a given set of objectives, generate outputs such as predictions, recommendations, or decisions influencing real or virtual environments.

Artificial Intelligence Task Force. In 2024, the Legislature established the AI Task Force, administered by the Office of the Attorney General. The AI Task Force is tasked with assessing current uses and trends of AI and making recommendations to the Legislature regarding guidelines and potential legislation for the use of AI systems. 

Provenance Data. A covered provider, to the extent commercially and technically reasonable, must include provenance data in images, video, audio, or content that is a combination thereof, that has been created or materially altered by the AI system. A covered provider must use commercially and technically reasonable methods to make the provenance data difficult to remove.   These requirements do not apply to video games, interactive experiences, and systems used solely for upscaling, noise reduction, or compression.

Government Disclosure. A government agency that makes available an AI system intended to interact with consumers must provide certain disclosures to each consumer, before or at the time of interaction, that the consumer is interacting with an AI system.

Enforcement. Enforcement is provided by the Attorney General under the Consumer Protection Act. A violation of the act is a civil violation and fines must not exceed more than $100,000 per covered provider.

Covered provider is defined as a person or entity that creates, codes, or otherwise produces a generative AI system that has over 1 million monthly users and is publicly accessible within the state to consumers for personal use.  It does not include state, local, or tribal government.

• Provenance Data.  The requirements do not apply to state, local, or tribal governments, or video games, interactive experiences, and systems used solely for upscaling, noise reduction, or compression.
• Enforcement.  Enforcement of the act by the Attorney General is performed under the Consumer Protection Act. 

• Redefines covered provider to mean a person or entity with more than 1 million monthly users within Washington.
• Strikes multiple definitions including personal provenance data, latency, and manifest.
• Eliminates requirements associated with making a provenance detection tool available to users and latent disclosure requirements.
• Changes requirements for provenance data to specify that, to the extent commercially and technically reasonable, a covered provider must take reasonable steps to ensure the provenance data is difficult to remove or tamper with.
• Specifies that nothing in the act requires the disclosure of a trade secret and circumstances where the act does not apply.
• Provides for enforcement exclusively by the attorney general and allows a civil penalty of up to $100,000 per covered provider.
• Adds a requirement that government agencies disclose to consumers when they are interacting with AI.
• Modifies the effective data to February 1, 2027.

The committee recommended a different version of the bill than what was heard. PRO: This bill is intended to focus on large users. The scope is limited to high risk products like data and photos. Some people will say this is not technically feasible. AI developers already have the tools required under this bill available. This approach closely appears the version in California that has already passed. This bill is narrow and practical. The bill protects workers from digital impersonation. When deepfakes spread it can jeopardize jobs before people have a chance to respond. We think the bill is completely technically feasible to implement. In fact, people have already begun implementing the California law. This bill creates consistency. This is the kind of thing we do in the consumer protection world all the time and it makes sense to do it here. It helps people distinguish authentic content from manufactured. This is not a government restriction on content. It's a disclosure of truthful content. This assures that the bad actors can't hide behind anonymity of algorithms.

CON: The bill as written is unworkable. The revised language introduces ambiguity for deployers. It introduces new terms that are undefined. The California bill is currently stalled and unworkable. The bill still has significant challenges. Our concern is timing and uncertainty. Stakeholders in California are still working through technical uncertainty. We don't know if California will translate to Washington. This bill can be compared to requiring a nutrition label. This bill instead requires the governments message to be embedded inside user speech.

The committee recommended a different version of the bill than what was heard. PRO:  This draft is focused on large, publicly accessible providers and adopts commercially and technically reasonable standards, supporting common commercial frameworks in place.  The delayed effective date gives a runway for developers to comply.  It will make innovation workable in Washington.  We support the recommended change to give the Attorney General more investigative tools.  AI-generated content is everywhere, and there are massive losses from fraud and harms caused by deepfakes.  People deserve to know when content is created by a machine.

OTHER:  WaTech doesn't have the resources to meet the current technical requirements in the bill, which will result in significant fiscal impact.  A simple fix would be to exempt state agencies from the definition of "covered provider."  The Attorney General will have increased enforcement costs with no private right of action unless investigative tools and the ability to recover costs that were in previous drafts are restored.