Whether you're looking for an angel investor, a growth advisor, or just want to connect — I'm always open to great ideas.
Get in TouchAI, startups & growth insights. No spam.
TL;DR: Washington State passed three significant AI safety bills before its 2026 legislative session adjourned — HB 2225 (chatbot disclosures and self-harm protections for children), HB 1170 (AI procurement disclosure and bias auditing), and SB 5395 (AI in health insurance coverage decisions). The moves come as states across the country race to fill a regulatory vacuum left by federal inaction, with Utah passing nine AI-related bills and Florida advancing an AI Bill of Rights through its Senate.
Washington State's legislature wrapped up its 2026 session with a significant cluster of AI-focused legislation that places the state among the most aggressive regulators of artificial intelligence in the country. Three bills cleared the chamber and are now headed to Governor Bob Ferguson for signature.
The timing is notable. Washington's legislative calendar forced a compressed final-week push on AI policy, and lawmakers chose to prioritize consumer protection over the more cautious approach some states have taken. The session closed with HB 2225, HB 1170, and SB 5395 each passing with substantial bipartisan support, according to reporting from Transparency Coalition AI's legislative update for March 13, 2026.
Washington is not starting from scratch on AI governance. The state already has some of the country's strongest data privacy laws through the Washington Privacy Act, and the new AI bills slot into that existing framework. Legislators argued that the surge in conversational AI tools — chatbots embedded in healthcare portals, school platforms, and social media — created specific harms that existing privacy law did not cover adequately.
The bills address three distinct domains: child safety and chatbot behavior, government and enterprise AI procurement transparency, and AI-assisted decision-making in health insurance. Together they represent the broadest AI regulatory package Washington has passed to date, and the combination of all three in a single session marks a deliberate attempt to set a state-wide baseline before federal standards arrive — if they ever do.
Governor Ferguson has not publicly opposed any of the three measures, and his office has previously expressed support for evidence-based AI oversight.
House Bill 2225 is the most consumer-facing of the three measures. It establishes disclosure requirements for chatbots interacting with minors and mandates specific safety protocols for conversations that touch on self-harm, eating disorders, or suicide.
Under HB 2225, any AI chatbot deployed on a platform accessible to users under 18 must clearly identify itself as a non-human system at the start of an interaction. The requirement applies to general-purpose AI assistants, companion apps, and embedded chatbots in educational software. The disclosure must be presented in plain language, not buried in terms of service.
The self-harm provisions are the most technically demanding part of the bill. When a chatbot detects conversation patterns associated with suicidal ideation or self-harm, it must follow a defined intervention protocol: surface crisis resources (including the 988 Suicide and Crisis Lifeline), refrain from providing detailed information that could facilitate harm, and — on platforms where the user is a verified minor — notify a designated responsible adult if the platform has parental oversight features enabled.
This is partly a response to documented cases involving chatbot companions like Character.AI, where multiple families have alleged that extended interactions with AI persona products contributed to self-harm incidents in teenagers. The Florida lawsuit involving a minor's suicide after interactions with a Character.AI persona drew national attention in 2024 and accelerated state-level legislative responses.
HB 2225 also limits what chatbots on children's platforms can do when users express emotional distress. Platforms will be prohibited from using vulnerable emotional states as engagement signals — that is, a chatbot cannot respond to sadness or anxiety in ways designed to prolong the conversation rather than address the user's wellbeing.
Enforcement falls to the Washington Attorney General's office. Violations carry civil penalties with a per-incident structure, and the AG can seek injunctive relief against non-compliant platforms.
House Bill 1170 takes a different angle, targeting how public agencies and large enterprises in Washington deploy AI systems. The bill establishes a disclosure regime for AI use in consequential decisions and requires bias auditing for high-stakes applications.
Under HB 1170, any state agency procuring or deploying an AI system for decisions that affect residents (hiring, benefits eligibility, permit approvals, child welfare assessments) must publicly document the system's purpose, the training data used, the vendor responsible, and any known limitations. These disclosures go into a centralized public registry maintained by the Washington Technology Solutions office.
The bias auditing requirement is more far-reaching. AI systems used in housing, employment, education, and lending decisions — whether deployed by public or private entities above a revenue threshold — must undergo third-party bias audits before deployment and at defined intervals thereafter. The audits must test for disparate impact across protected classes and the results must be filed with the state.
| HB 1170 requirement | Applies to | Frequency |
|---|---|---|
| Procurement disclosure | All state agencies | Per contract |
| Use-case documentation | State agencies + enterprises | Annual |
| Third-party bias audit | Housing, employment, lending, education AI | Pre-deployment + biennial |
| Disparate impact report | All audited systems | Biennial |
This structure mirrors elements of New York City's Local Law 144, which requires bias audits for automated employment decision tools. But Washington's scope is broader: it covers multiple sectors rather than just employment, and it captures both public agency AI and private-sector deployments above the size threshold.
Critics from the technology industry have raised concerns about the cost and complexity of third-party auditing, particularly for smaller firms. Proponents argue that without a standardized audit requirement, companies have no financial incentive to address algorithmic bias voluntarily.
Senate Bill 5395 may be the most consequential of the three measures in terms of economic impact. It directly addresses how health insurers in Washington can use AI systems to make coverage and claims decisions.
The bill was motivated by a growing body of evidence that health insurers are deploying AI tools that deny claims at higher rates than human reviewers, often based on population-level statistical models rather than the specific circumstances of individual patients. A 2024 investigation into major payers found that AI-assisted prior authorization systems had denial rates two to three times higher than peer-reviewed medical standards would suggest is appropriate.
Under SB 5395, health insurers must disclose when AI or algorithmic systems are used in coverage decisions, prior authorization, and claims processing. The disclosure must reach the patient and the treating provider at the time of the decision — not retroactively.
The bill also establishes an explainability requirement. When a claim or authorization is denied by or with AI assistance, the insurer must provide a human-readable explanation of the specific factors that contributed to the denial. Generic references to "clinical criteria" or "medical necessity guidelines" without specific application to the patient's case do not satisfy the requirement.
Most significantly, SB 5395 creates an AI-assisted appeal right. Patients denied coverage through an AI-assisted process have the right to a human clinical reviewer for their appeal. The human reviewer must be a licensed clinician in the relevant specialty, and the appeal timeline is shortened relative to standard review. Insurers cannot substitute a different AI system for the appeal review.
The health insurance industry lobbied against the bill, arguing that AI actually speeds up legitimate claim approvals and that the added disclosure and explainability requirements will increase administrative costs without improving outcomes. Supporters countered that patient harm from opaque AI denials is documented and that transparency requirements are a minimum standard for systems making life-affecting medical decisions.
Washington is not operating in isolation. Across the country, state legislatures that were largely reactive to AI in 2023-2024 are now actively setting rules.
Utah had one of the most productive AI legislative sessions in the country, passing nine AI-related bills in a single short session. The Utah package covers a range of topics:
Utah's approach is notable for its specificity. Rather than broad AI governance frameworks, the state has passed targeted bills aimed at the specific harms that AI tools are already causing. This issue-by-issue strategy allows faster legislative action than comprehensive AI governance bills, which tend to stall in committee over definitional disputes.
| State | 2026 AI bills passed | Primary focus areas |
|---|---|---|
| Washington | 3 | Chatbot safety, bias auditing, health insurance |
| Utah | 9 | Deepfakes, voice cloning, advertising, companion apps |
| Colorado | 2 | AI in hiring, automated decision systems |
| Illinois | 1 | AI hiring interview recording disclosure |
| Texas (pending) | 4 | AI in education, child safety |
The pace of state-level AI legislation has accelerated sharply from 2024 to 2026. The National Conference of State Legislatures tracked 135 AI-related bills introduced across state houses in 2024. That number is on track to exceed 400 in 2026.
Florida has taken a more comprehensive approach. Senate Bill 482, branded as the Florida AI Bill of Rights, passed the Florida Senate 35-2 — a margin that signals unusually strong bipartisan consensus for a technology regulation bill.
The Florida bill establishes baseline rights for state residents interacting with AI systems across both public and private sectors. Its core provisions include the right to know when AI is used in a consequential decision, the right to appeal an AI-assisted decision to a human reviewer, the right to data portability regarding AI-generated profiles, and prohibitions on specific AI applications in law enforcement (predictive policing tools, facial recognition without warrant in most contexts).
Florida Governor Ron DeSantis has historically been skeptical of technology regulation, favoring market-based approaches. The 35-2 Senate vote creates political pressure to sign the bill, but his office has not committed to a timeline. The bill now moves to the Florida House, where its version has been more contentious.
What makes the Florida bill significant beyond its content is its framing. Branding AI regulation as a "Bill of Rights" mirrors the rhetorical strategy consumer advocates have used in other policy fights — data privacy, internet neutrality — to build broad popular support that makes legislative opposition politically costly.
If signed, the Florida AI Bill of Rights would cover a state of approximately 22 million people and represent the most sweeping single-state AI governance measure enacted in the United States.
The proliferation of state AI laws creates a structural challenge for companies operating nationally. Unlike federal law, which sets a single floor, state AI regulation creates a patchwork of requirements that vary by jurisdiction in ways that are difficult to harmonize.
Consider the chatbot disclosure requirement alone. Washington's HB 2225 requires disclosure at conversation initiation for minors. California's proposed AB 2602 requires disclosure for any AI-generated media. Utah's disclosure laws apply to political contexts. None of these requirements use identical definitions, triggers, or enforcement mechanisms.
For a company deploying a general-purpose chatbot in all 50 states, compliance requires:
The compliance overhead scales with the number of state laws. At 9 laws in 2026, compliance is manageable for large companies. At 40+ laws — a plausible scenario by 2028 at current legislative velocity — the burden becomes prohibitive for mid-size companies and creates a competitive advantage for large platforms with dedicated legal and engineering resources.
This fragmentation is not a bug from the states' perspective. Many legislators are explicitly using state action to pressure the federal government to pass preemptive national standards.
Technology companies are taking varied approaches to the emerging state AI regulatory landscape. The broad categories are: compliance-forward, advocacy-led, and geographic risk management.
Compliance-forward companies are treating state AI laws as permanent structural requirements and building compliance infrastructure accordingly. Microsoft, Google, and Amazon have each expanded their AI policy and legal teams significantly. Google's AI governance team reportedly tripled in size in 2025. These companies are building internal tooling to track state AI legislation, map requirements to specific products, and generate compliance documentation at scale.
Advocacy-led companies are investing in lobbying to shape state legislation before it passes, and in some cases to preempt it at the federal level. The major AI labs (OpenAI, Anthropic, xAI) have all expanded their Washington, D.C. policy presence and are engaging state capitals in ways they were not in 2023-2024. The goal is to shape bills that create workable requirements while avoiding provisions that would materially change product behavior.
Geographic risk management is primarily a strategy for smaller AI companies. Rather than building compliance infrastructure for every state, some companies are choosing to initially avoid deploying consumer-facing AI products in highly regulated states. This creates accessibility gaps — users in states with strong AI regulation may find that certain AI services simply are not available to them.
The insurance sector response to SB 5395-style laws is more constrained. Health insurers cannot easily withdraw from state markets, so the Washington bill will require material changes to how AI is deployed in claims processing by all payers operating in the state.
The acceleration of state AI legislation is a direct consequence of federal inaction. At the federal level, comprehensive AI regulation has stalled across multiple Congresses.
The American AI Act, introduced in 2024, has not cleared committee. The Algorithmic Accountability Act, which would have required impact assessments for consequential AI systems, similarly failed to advance. The White House's AI Executive Order from October 2023 directed agencies to develop voluntary commitments and sector-specific guidance, but created no enforcement mechanism or statutory baseline.
The Biden administration's actions were largely process-focused. The Trump administration, which returned to office in January 2025, has emphasized AI deregulation and American AI competitiveness, explicitly pulling back from some Biden-era AI oversight initiatives.
In this federal vacuum, states are not filling a temporary gap. They are building the permanent regulatory architecture. When and if federal AI legislation passes, it will either preempt state laws (as some industry groups want) or establish a federal floor above which states can add additional requirements (as most state legislators prefer). The current trajectory strongly favors the floor model, because the political coalition for preemption — industry groups and federal agencies — is weaker than the coalition against it — state AGs, consumer advocates, and state legislators who do not want their work negated.
The European Union's AI Act, which has been in force since August 2024, provides a reference model that several U.S. state bills have drawn on. Washington's bias auditing requirements in HB 1170 are structurally similar to the EU AI Act's conformity assessment requirements for high-risk AI systems.
The March 2026 legislative session in Washington, combined with Utah's nine-bill package and Florida's AI Bill of Rights vote, marks a qualitative shift in how AI regulation is developing in the United States.
For the first two years of the generative AI era (2022-2024), state legislatures largely watched. Bills were introduced but rarely passed. The regulatory energy was more performative than substantive. That changed in 2025, when consequential harms from AI — denied insurance claims, self-harm incidents involving AI companion products, AI-generated disinformation in elections — became sufficiently documented that legislators had specific problems to solve rather than abstract futures to regulate.
The implications for companies building or deploying AI systems in the United States are significant:
Washington's three bills will take effect on staggered timelines after the Governor's signature, giving companies 6-18 months to come into compliance depending on the specific provision. The clock is running.
Washington State passed HB 2225 (chatbot safety requirements for minors, including self-harm intervention protocols), HB 1170 (AI disclosure requirements for government procurement and bias auditing for high-stakes AI), and SB 5395 (AI transparency and appeal rights for health insurance coverage decisions).
HB 2225 requires AI chatbots accessible to minors to disclose their non-human nature at the start of interactions, follow defined safety protocols when conversations involve self-harm or suicidal ideation, and surface crisis resources. Platforms cannot use vulnerable emotional states to increase user engagement.
HB 1170 requires third-party bias audits for AI systems used in housing, employment, education, and lending decisions. Audits must test for disparate impact across protected classes and must be filed with the state before deployment and every two years thereafter.
SB 5395 requires health insurers to disclose when AI is used in coverage or claims decisions, provide human-readable explanations for AI-assisted denials, and grant patients the right to a human clinical reviewer for appeals of AI-assisted denials.
Utah passed nine AI-related bills in a single short session, covering topics including AI-generated political advertising, voice cloning fraud, fake reviews, AI companion platforms for minors, and liability frameworks for AI-assisted professional advice.
Florida Senate Bill 482, branded as the Florida AI Bill of Rights, passed the Florida Senate 35-2. It establishes the right to know when AI is used in consequential decisions, the right to human appeal, data portability rights, and restrictions on AI in law enforcement. The bill awaits a House vote and the Governor's signature.
Federal AI legislation has stalled repeatedly in Congress, and the Trump administration has prioritized AI deregulation. States are filling the regulatory vacuum by passing targeted bills addressing specific documented harms, including chatbot self-harm incidents, AI-assisted insurance denials, and algorithmic bias in employment and housing.
Companies operating nationally must track varying disclosure, auditing, appeal, and enforcement requirements across multiple states. Requirements use different definitions and triggers. For large companies, compliance is costly but manageable. For mid-size companies, the fragmented patchwork creates significant overhead that favors large platforms with dedicated legal teams.
Federal AI preemption is unlikely in the near term given the political environment. Most state legislators and consumer advocates prefer a federal floor model that lets states add requirements above a national baseline, rather than full preemption. Companies should not count on federal law eliminating state AI requirements.
Washington's bias auditing requirements in HB 1170 are structurally similar to the EU AI Act's conformity assessment requirements for high-risk AI systems. The EU AI Act has been in force since August 2024 and has served as a reference model for several U.S. state bills.
The EU Competition Commissioner targets Nvidia, Meta, and Google in an unprecedented antitrust investigation spanning the entire AI value chain from chips to deployment.
Anthropic escalates its showdown with the Pentagon, seeking an appeals court stay after being designated a supply-chain risk for refusing unrestricted military AI access.
The FDA's breakthrough designation for RecovryAI signals a shift from AI diagnostics toward AI-driven patient compliance and post-op care.