SHB 2157
In CommitteeHouse
High-risk AI
Regulating high-risk artificial intelligence system development, deployment, and use.
This status may be delayed. See Action History below for the latest updates.
How does a bill become law?
- Introduced: The bill is filed and assigned a number.
- Committee: A subject-matter committee holds hearings, takes public testimony, and decides whether to advance the bill.
- Floor Vote: The full chamber (House or Senate) debates and votes on the bill.
- Opposite Chamber: The bill repeats the committee and floor vote process in the other chamber.
- Governor: The Governor reviews the bill and decides whether to sign or veto it.
- Signed: The bill has been signed into law.
AI Analysis
This bill regulates the use of high-risk artificial intelligence systems in Washington to prevent algorithmic discrimination and ensure transparency. It applies to AI used for consequential decisions—like hiring, housing, or legal outcomes—and requires developers and deployers to disclose how the systems work, assess risks, and protect consumers.
- Creates a new legal category of 'high-risk artificial intelligence systems'—AI used to make 'consequential decisions' like employment, housing, lending, education, or legal outcomes.
- Requires developers to disclose intended uses, limitations, and risks of algorithmic discrimination, and to provide documentation to deployers—including system cards and predeployment impact assessments.
- Mandates that deployers implement a risk management policy, complete impact assessments before using high-risk AI, and disclose AI use to consumers—including how the system works and what data it uses.
- Prohibits algorithmic discrimination and requires developers and deployers to use 'reasonable care' to prevent it; compliance with national or international AI risk management standards (e.g., NIST or ISO/IEC 42001) creates a legal presumption of compliance.
- Requires identification of synthetic content (e.g., AI-generated images or videos) in a way that is accessible to consumers, with limited exceptions for artistic or clearly labeled works.
- Grants consumers the right to sue for violations, with courts able to issue injunctions and award attorneys’ fees; developers or deployers can avoid liability if they cure violations within 45 days of discovery.
Who is affected
- Developers of high-risk AI systems — Businesses that develop or significantly modify high-risk AI systems and earn over $100,000 annually in Washington must provide documentation, conduct impact assessments, and ensure their systems do not cause algorithmic discrimination.
- Deployers of high-risk AI systems — Businesses that use high-risk AI to make consequential decisions (e.g., hiring, housing, lending, education, or legal decisions) must implement risk management policies, complete impact assessments, disclose AI use to consumers, and provide explanations for adverse decisions.
- Consumers — Washington residents making personal, household, or consumer-related decisions (e.g., applying for jobs, housing, loans, or education) gain rights to know when AI is used, how it affects them, and how to challenge decisions.
- Insurers — Insurance companies operating in Washington may be exempt if they are regulated by the Office of the Insurance Commissioner and subject to existing rules on discrimination in insurance practices.
- Healthcare providers and health systems — Healthcare providers and telemedicine services may be exempt if their AI use falls under HIPAA or state health privacy laws and is for administrative or clinical decision support purposes.
Pro/Con Analysis
Stronger case for benefits
Potential Benefits (5)
Grants consumers explicit rights to know when AI is used in consequential decisions and to receive clear explanations — empowering individuals to challenge biased outcomes and hold deployers accountable.
Rights & LibertiesPeopleRef: Sec. 3(4), (5)Creates a private right of action with attorney’s fees and injunctive relief, lowering barriers for individuals to challenge algorithmic discrimination — especially important where harms are systemic but individually subtle.
Rights & LibertiesPeopleRef: Sec. 2(1), Sec. 3(1), Sec. 5Requires synthetic content identification and predeployment impact assessments — helps prevent deepfakes and biased decision-making in housing, employment, and legal contexts, protecting vulnerable populations.
Public SafetyPeopleRef: Sec. 2(2)(c), Sec. 3(3)(b)(vii)Mandates documentation of data categories, mitigation steps, and postdeployment monitoring — increases accountability and enables civil rights enforcement, especially in high-stakes domains like hiring and lending.
Business & EmploymentPeopleRef: Sec. 3(3)(b)(ii), (viii), (ix)Explicitly covers AI use in education enrollment and requires plain-language disclosure — helps prevent algorithmic bias in college admissions, financial aid, or discipline, protecting low-income and first-generation students.
EducationPeopleRef: Sec. 1(3), Sec. 2(7), Sec. 3(4)
Potential Concerns (5)
Mandates that deployers disclose AI use and decision reasoning to consumers, but only *after* a decision is made — not before — limiting consumer ability to meaningfully opt out or challenge use in real time.
Rights & LibertiesPeopleRef: Sec. 3(4), (5)Exemptions for law enforcement and national security (Sec. 4(1)(j), (k)) and lack of audit or oversight provisions for those uses create risk of unchecked AI in policing, surveillance, or immigration enforcement — especially harmful to marginalized communities.
Public SafetyLean peopleRef: Sec. 3(3)(b)(vii), Sec. 3(4)(e)(v)Allows reliance on existing regulatory frameworks (e.g., NIST, ISO, insurance rules) as compliance shortcuts, which may be outdated, incomplete, or less rigorous than the bill’s own standards — weakening enforcement and creating regulatory arbitrage.
Business & EmploymentLean peopleRef: Sec. 3(3)(c)(ii)Cures violations within 45 days without penalty — creates a “use first, fix later” loophole that incentivizes deployment of risky systems and delays accountability, especially for harms that compound over time (e.g., housing discrimination).
Rights & LibertiesPeopleRef: Sec. 5(2)(b)Broad exemptions for insurers and healthcare providers allow use of high-risk AI in critical areas (e.g., insurance underwriting, clinical decision support) with minimal transparency or consumer recourse — disproportionately impacts low-income and chronically ill residents.
Business & EmploymentPeopleRef: Sec. 4(8), (11)
Who Is Most Affected
Low- and moderate-income Washingtonians applying for jobs, housing, or credit benefit significantly — the bill provides tools to detect and challenge biased AI decisions that disproportionately affect marginalized groups.
Small and mid-sized businesses deploying AI for hiring, lending, or customer service face new compliance costs (impact assessments, documentation, training), but avoid liability if they follow the risk management framework — net effect is mixed but manageable for most.
Large tech platforms and SaaS providers face significant compliance burdens, but their scale allows them to absorb costs more easily than smaller firms — and they benefit from exemptions and safe harbors for existing standards.
Insurance and healthcare sectors are largely exempt, preserving their current AI practices — reducing regulatory burden for these industries but limiting consumer protections in critical sectors.
Law enforcement and state/federal agencies gain broad exemptions for security and surveillance uses — enabling continued use of high-risk AI in policing and immigration without transparency or oversight.