WA AIR Policy Framework

Why We Need To Take on AI

Over the last two decades, Big Tech has rolled out product after product with virtually no oversight or accountability. Their “move fast and break things” model has caused serious societal harm while they have flooded local, state and federal governments with campaign contributions to stop regulation before it can start.

As of 2026, one in four federal lobbyists represent the AI industry, which is increasingly a clear and present danger to democracy, privacy and our world.

Building on 2026 Momentum

In the 2026 legislative session, multiple bills made significant progress despite the short session. While bills to regulate data centers (HB 2515) and surveillance pricing (HB 2481) were unable to overcome the combination of industry opposition and a crowded calendar, advocates were able to muster significant public support.

With time to build a real coalition and mobilize voters, plus the full-length session, we believe Washington can make meaningful progress in 2027.

Public Opinion Favours Regulation

Polls show that a growing, cross-partisan majority of voters is deeply concerned about the rapid growth of AI. A wide, cross-partisan majority supports strong regulation of AI as well as adoption of general privacy regulations. With the Trump administration and much of the GOP fully captured by the industry, the burden falls on states to take regulatory action.

How Washington Can Lead

Washington is uniquely positioned to model a “third way” of AI governance that preserves the capability to innovate while protecting already strained public resources. Washington can create a unified regulatory front that resists federal preemption and protects the powers of the state to regulate consumer safety and state resources.

Data & Biometric Privacy

HIPAA presents significant gaps, also not covered by WA “My Health My Data,” which lead to lapses whereby Washington residents cannot control their privacy:

• HIPAA “anonymized” healthcare data can be deanonymized using AI, revealing medical conditions of vulnerable individuals.
• Data stored according to HIPAA rules is not regulated for accuracy, and medical chatbots are considered by physicians to be a top public health hazard in 2026.
• Employers penalizing workers for low output, tracked with AI, may be measuring metrics related to medical conditions, with workers having no recourse.

Similarly, both FERPA and the WA “Student User Privacy in Education Rights” act do not cover:

• That algorithmic advertising is banned but not predictive analytics, flagging “at risk” students in ways that they cannot object to or control well into adulthood.
• “School officials,” via a loophole, can expose intellectual property of minors to train AI models with neither student nor parental consent.
• The lack of recourse for students from marginalized communities who cannot opt out of AI disproportionately flagging them for academic integrity and behavioural concerns.

Algorithmic Guardrails

Current consumer protection laws do not account for the phenomenon of “surveillance pricing” where AI algorithms adjust the price of essential goods like groceries in real time, based on an individual’s inferred willingness to pay. When applied to wages, as it is used for much of “gig economy” labour, it becomes a vector for rampant wage discrimination that the state cannot effectively regulate.

Further, there are no state-level mandates that require or guarantee that a human has reviewed high-stakes decisions in healthcare, criminal justice, lending, or employment.

Public Infrastructure Transparency

Current state law does not require data center operators to disclose via audited third party, their water consumption, carbon emissions, or energy mix to the communities in which they operate. The use of nondisclosure agreements (NDAs) are prevalent and yet unregulated. Communities often discover, years after project completion, that:

• Consumer electricity prices exceed headline CPI inflation
• They experience more than a 25% increase in electricity prices over 5 years
• Data centers can use more than 10% of an entire county’s water consumption allotment per day, necessitating in one case $212,000,000 in water infrastructure grants
• Data centers contribute significantly to air pollution, the effects of which result in chronic illnesses estimated to cause $20,000,000,000 of burden to public health systems

Economic Fairness

The federal and state tax codes are currently “automation biased,” incentivizing the replacement of human labor with capital-intensive AI systems. Companies can frequently deduct 100% of their investment in hardware and software while human payroll remains heavily taxed. This creates a fiscal “death spiral” where the tax base shrinks even as demand for social safety nets increases.

If AI automates 1 out of 10 jobs, worker safety net programs like PFML, WA Cares, and Unemployment Insurance will face immediate insolvency totalling more than $4,000,000,000 in impact. Furthermore:

• AI systems are already being used to suppress labour organizing, profit from work outputs where workers cannot otherwise reap collective benefits, and devalue labour.
• Under current regulations, the “Developer Liability Shield” (analogous and complementary to Section 230) diminishes the legal incentive for AI providers to implement downstream safeguards.

Economic Fairness

Automation Impact Levy

Washington may leverage the existing framework of the “Advanced Computing B&O” levy via an “Automation Impact Fee” on commercial AI deployments, scaled by usage volume, exempting nonprofits, academic research, and small businesses. Using a localized capture model of a reported $650 billion in capital expenditure across the AI sector, with a 10% automation surcharge, the state could recover up to $700M a year in revenue. At a 20% surcharge rate (~$1.4B per annum), this could fund a general wage insurance program covering:

• Up to 80% of the wage gap from workers who accept lower-wage work after displacement, up to $40,000 (~$500M/yr est.)
• Short-term wage displacement stipends of up to $10,000 one time (~$120M/yr est.)
• Universally accessible retraining vouchers and grants for public educational institutions up to $10,000 per person (~$500M/yr est.)
• Portability of wage insurance benefits, tying them to the individual and not the job

Profit-Based Business Tax Models

The state must also shift its tax base to capture the “supernormal” profits of the AI industry, including exploring a state-level Value Added Tax (VAT) / Business Activity Tax (BAT) and a “high automation business” additional B&O surcharge.

Data & Biometric Privacy

Prohibited Technology

Washington must enact a categorical ban on the deployment of facial recognition technology in all consumer goods. Additionally, parallel to the EU AI Act, we must ban: social scoring systems, surveillance-based pricing, subliminal manipulation, and employee surveillance (including the use of AI to suppress labour organizing). Legislation should also prohibit the “passive harvesting” of data by mass surveillance systems like Automated License Plate Readers (ALPRs).

Digital Rights Act

The state should implement a comprehensive Digital Rights Act codifying European-style data subject rights, including:

• Right to Erasure — “Machine unlearning” where personal data must be removed from model training weights upon request
• Right to Review and Rectification — An obligation to provide a mechanism for Washington residents to correct data processed by AI and resulting decisions
• Right to Object — An absolute right for residents to opt out of data collection for automated decisioning in credit, lending, predictive analytics, targeted advertising, employment, healthcare, education, housing, and government service determinations
• Right to Accountability — AI developers and deployers bear legal liability for decisions made by their autonomous systems
• Data Portability — Mandate for companies to provide data in machine-readable formats (JSON/XML)

Institutional Guardrails

The state must extend HIPAA-level protections to all entities processing health-related data. AI deployments in the workplace should be grounds for mandatory collective bargaining, guaranteeing workers the right to bargain around conditions of AI usage in the workplace freely and in good faith.

Algorithmic Guardrails

Data Lineage Disclosure

Developers must provide a “Data Lineage Disclosure” for all AI models commercially deployed in Washington, including a summary of training data origin and ownership, documentation of bias mitigation efforts, and publicly available “Model Cards.” This framework aligns with CA AB 2013.

Right to Authenticity

The state must establish a “Right to Authenticity,” requiring that all AI-generated images, audio, and video include persistent, machine-readable metadata identifying the content as synthetic, plus a human-visible watermark. Residents also possess a “Right to Interact with a Human” in any interaction with an AI chatbot representing a licensed professional.

Public Infrastructure Improvements

Public Records Reform

Washington should ban the use of nondisclosure agreements (NDAs) between data center developers and government agencies. Public records laws must be reformed to ensure all data center permit applications, water rights transfers, and energy usage projections are subject to immediate and full public disclosure.

Tribal Consent

Washington should guarantee that no large-scale energy or industrial project shall proceed without meaningful consultation and the free, prior, and informed consent of affected Tribal Nations, particularly regarding impacts on treaty-protected water rights and salmon recovery.

Data Center Sustainability Reporting

Each “Emerging Large Energy Use Facility” (ELEUF) — defined as a facility with a demand of 20 megawatts or more — must publish an annual sustainability report covering water usage (WUE), energy usage (PUE), grid impact, and Scope 1, 2, and 3 emissions.

Commitment to Sustainable Grids

Small Modular Nuclear Reactors (SMNRs) should be categorically unpermitted due to unproven commercial viability, high-level waste concerns, and opposition from Tribal Nations. Mature, cost-effective technologies like wind, solar, and long-duration storage should be promoted instead.

Development Moratorium

Washington should enact a statewide moratorium on new data center development until new legislation has ensured renewable energy buildout, watershed protections, and annual sustainability reporting can be guaranteed. Moratoria should have expected expiry dates adjustable upon evidence of long-term community commitment and project funding guarantees.

Executive Accountability

The state must establish clear criminal and civil liability for executives of companies whose products cause harm to Washington residents, including:

• Establishing “mens rea” (recklessness or indifference) for executives who deploy agentic AI systems that cause critical infrastructure failures or large-scale physical harm
• Adopting frameworks like the Colorado AI Act, establishing a “Duty of Care” standard for AI developers
• Strengthening the Consumer Protection Act to allow fines of the greater of $100,000 or 5% of worldwide annual gross revenue, per diem, for ongoing violations
• Enabling a private right of action, as in HB 2225, so that the state attorney general is not the bottleneck to preventing just and fair outcomes
• Requiring developers to fund a “Decommissioning Bond” to ensure that defunct data centers are transitioned or dismantled without leaving an ecological or financial burden on the host community

The Interstate Compact

To circumvent the “America’s AI Action Plan” and its associated federal preemption threats, Washington should form an “AI Governance Interstate Compact” with Colorado, California, and other like-minded states. This compact would create a unified legal theory defending state authority under the Tenth Amendment and the Dormant Commerce Clause. This “Persuasive Resistance” strategy would signal a unified state position to Congress and provide a collective defense against DOJ litigation.

Third Party Safety Audits

Compliance with the Washington AI Bill of Rights should be a condition for business licensure. The state will establish a “Center for AI Safety and Evaluation” to oversee third-party audits based on the NIST AI Risk Management Framework and ISO/IEC 42001 standards.