Washington State AI Law SB 5838: What Businesses Need to Know (2026)

What Is SB 5838?

Washington State SB 5838 is part of the wave of state-level AI legislation that accelerated across the US after Colorado passed SB 205 in 2024. Like the Colorado law and the EU AI Act that inspired them, Washington's approach is risk-based: more requirements attach to AI systems used in higher-stakes decisions.

The law focuses on "algorithmic decision tools" — software that uses AI, machine learning, or statistical modeling to substantially inform or make decisions with significant effects on individuals. The key organizing concept is the "consequential decision": a decision that has a significant effect on an individual's access to education, employment, financial services, healthcare, housing, insurance, or similar high-stakes domains.

Who Is Covered

SB 5838 creates obligations for two categories of entities:

Developers

Companies that design, develop, and commercialize AI systems intended to be used in consequential decisions affecting Washington residents. Developer obligations focus on documentation and transparency: providing deployers with information sufficient to assess the system's capabilities, limitations, and potential for discrimination.

Deployers

Companies that deploy or use AI systems to make or substantially inform consequential decisions affecting Washington residents. Deployer obligations are more extensive and include the individual-facing disclosure, opt-out, and impact assessment requirements. If your company uses a third-party AI tool in employment screening, credit decisions, or housing applications, you're likely a deployer.

Note: Many companies are simultaneously developers (they build AI features) and deployers (they use those features in their own operations). Both sets of obligations apply.

Covered Consequential Decisions

The law covers AI-assisted decisions in these domains when they significantly affect Washington residents:

• Employment: Hiring, promotion, termination, compensation, and performance evaluation decisions, including resume screening, candidate ranking, interview analysis, and productivity monitoring
• Credit and financial services: Loan applications, credit limits, interest rates, insurance underwriting and pricing, and similar financial decisions
• Housing: Rental applications, mortgage approvals, and real estate transactions
• Education: Admissions, placement, financial aid, and academic evaluation decisions
• Healthcare: Treatment recommendations, prior authorizations, and care management decisions

Core Compliance Requirements

1. Disclosure to Affected Individuals

When an AI system substantially informs a consequential decision, the deployer must disclose this to the individual affected. The disclosure must include:

• The fact that an algorithmic decision tool was used
• The purpose and role of the AI system in the decision
• The categories of data used as inputs
• Information about the individual's rights under the law

The disclosure must be plain-language, provided before or contemporaneously with the consequential decision, and not buried in terms of service. Consider how this affects your employment application workflows, credit decision notices, and insurance underwriting communications.

2. Right to Opt Out and Request Human Review

Individuals have the right to opt out of being subject to solely automated consequential decisions and to request that a human review the decision. Deployers must implement a process for handling these requests. The human review must be meaningful — not simply having a human rubber-stamp the AI's recommendation, but actually considering the individual's circumstances.

This mirrors GDPR Article 22 rights for EU residents and Colorado's analogous provision. If your employment screening process has no human review option, you need to build one.

3. Non-Discrimination and Algorithmic Fairness

Deployers may not use algorithmic decision tools in ways that result in unlawful discrimination based on protected characteristics. This is largely coextensive with existing anti-discrimination law, but the statute makes clear that using AI doesn't shield you from discrimination liability — if the AI produces discriminatory outcomes, the deployer is responsible.

For HR technology users in particular: AI hiring tools that produce disparate impact on protected groups create liability regardless of whether discrimination was intentional. The AI's "objectivity" is not a defense.

4. Impact Assessments for High-Risk Deployments

Deployers using AI in covered high-risk categories must conduct impact assessments before deployment and update them periodically (annually or when the system materially changes). Impact assessment components typically include:

• Description of the AI system, its purpose, and the decisions it informs
• Categories of personal data used as inputs
• Evaluation of potential discriminatory impacts by protected characteristic
• Technical and organizational measures implemented to mitigate bias and errors
• Results of accuracy and fairness testing
• Post-deployment monitoring processes

Impact assessments must be maintained and made available to Washington's Attorney General on request. They're not publicly filed but can be subpoenaed in litigation.

5. Developer Documentation Obligations

AI system developers must provide deployers with documentation that enables compliance, including:

• Description of the system's intended purpose and use cases
• Training data sources and data governance practices
• Performance metrics including accuracy, bias testing results, and known limitations
• Instructions for compliant deployment and ongoing monitoring

If you sell AI software or APIs used in Washington, review whether your documentation currently meets these requirements. Deployer customers may contractually require this documentation as part of their own compliance programs.

How Washington Compares to Colorado

Washington SB 5838 and Colorado SB 205 address the same problem space with similar approaches. Key differences:

Companies already compliant with Colorado SB 205 are well-positioned to extend that compliance program to Washington. The documentation, impact assessment, and disclosure frameworks carry over with Washington-specific adjustments.

The Growing Map of State AI Laws

Washington and Colorado are not isolated. The state AI legislation landscape as of mid-2026:

• Colorado SB 205: In effect; consequential decisions, impact assessments, developer/deployer framework
• Texas SB 2060 (TRAIGA): Algorithmic discrimination in covered sectors; Texas AG enforcement
• New York City Local Law 144: AI hiring bias audits; annual third-party audit required for covered tools
• Illinois AI Video Interview Act: Disclosure and consent for AI video interview analysis
• California SB-942: Disclosure requirements for AI-generated content

At least 12 additional states introduced AI bills in 2025–2026 legislative sessions. Companies with national operations should be building compliance programs that can be adapted state-by-state rather than starting from scratch for each new law.

Your Compliance Checklist for Washington SB 5838

Enforcement and Penalties

Washington SB 5838 is enforced by the Washington State Attorney General, not individual plaintiffs. There is no private right of action — individuals cannot sue you directly under the statute, though they may still have discrimination claims under Washington's Law Against Discrimination or federal civil rights laws.

The AG can investigate violations, issue civil investigative demands, seek injunctive relief, and impose civil penalties. Focus enforcement attention on developing genuine compliance practices rather than just documentation — AG investigations typically look at whether the documented policies are actually implemented.