Washington Governor Bob Ferguson signed House Bill 2225 on March 24, 2026, creating one of the most consequential state AI laws passed so far this year. The law specifically targets AI companion chatbots — systems designed to build ongoing emotional relationships with users — and it does something most state AI bills have avoided: it hands enforcement power directly to the people affected.

Under HB 2225, a violation of the law is treated as an unfair or deceptive act under Washington’s Consumer Protection Act. That is not just a regulatory sanction. It gives individual users the right to sue AI chatbot operators in civil court. The law takes effect January 1, 2027.

Washington is the home state of Microsoft, Amazon, and a significant share of the AI industry. A consumer protection lawsuit trigger in that state carries weight beyond its borders.

What the Law Actually Requires

The law applies to AI systems designed to simulate human conversation in a way that builds emotional connection or attachment — think AI companion apps, emotionally supportive chatbots, and similar products. Standard business chatbots that answer FAQ questions or process orders are in a different risk category, but the line between “customer service bot” and “emotionally engaging assistant” is blurring fast.

For all users:

Operators must provide a clear and prominent disclosure at the start of every interaction that the user is talking to an AI, not a person. That disclosure must repeat at least every three hours of continued interaction. The clock resets per session, but the requirement is ongoing for long interactions.

Operators must also establish and publicly disclose protocols for detecting and responding to users expressing suicidal ideation or self-harm. This is not a vague best-practice requirement. The law specifies that protocols must include: reasonable methods to detect expressions of suicidal ideation or eating disorders, responses that refer users to crisis hotlines and text lines, measures to prevent the chatbot from generating content that encourages self-harm, and annual public reporting of how many crisis referrals were issued through the app or service in the prior year.

That last item is significant. Most compliance requirements stay internal. Washington is requiring companies to publish data about how many times their chatbot triggered crisis referrals. That is accountability in a form that reporters, researchers, and regulators can see.

Additional requirements when minors are involved:

When an operator knows a user is a minor, or the product is directed at minors, two more requirements kick in. The operator must implement reasonable measures to prevent sexually explicit content or suggestive dialogue. And operators must prevent the chatbot from using manipulative engagement techniques that are specifically designed to prolong an emotional relationship with the user.

The last requirement is aimed squarely at the mechanics that make AI companions compelling — reward loops, simulated attachment, designed escalation of emotional dependency. Washington is the first state to explicitly target those design patterns in statute.

Why the Private Right of Action Changes Everything

Most state AI laws to date have been enforced by state attorneys general or regulatory agencies. HB 2225 changes the enforcement model.

By routing violations through the Consumer Protection Act, Washington has created individual lawsuit rights. A user who believes an operator violated the disclosure rules, failed to implement crisis protocols, or used manipulative techniques on them as a minor can take that to court without waiting for a government agency to act.

Tennessee created a similar private right of action in April for AI systems that claim to be licensed therapists. Washington’s law covers a broader category and adds the annual reporting requirement.

For companies with national reach, the combination matters: you cannot geofence your way around a consumer protection lawsuit. If Washington residents are using your product, HB 2225 applies.

The Companion Chatbot Market Is Not a Niche

The businesses in scope are not just Replika or character.ai. The companion chatbot design pattern has been spreading into mainstream applications: customer loyalty platforms, mental wellness apps, fitness coaching tools, educational assistants, customer service bots that are optimised for engagement. If the product is designed to make users feel heard and understood — and many modern AI tools are — the emotional bond question is worth examining.

What This Means for Business

Nine months until the law takes effect sounds like runway. It is not much for legal review, product changes, protocol design, disclosure implementation, and public reporting infrastructure.

The practical steps that Washington’s law demands:

First, audit every customer-facing AI product for disclosure compliance. Is the AI nature of the interaction clear at the start and through long sessions? The threshold for what counts as “clear and prominent” is not defined in the law, which means early litigation will set it. Better to be conservative now.

Second, establish crisis referral protocols if any product could receive mental health disclosures from users. The law applies broadly. A customer service chatbot that a user decides to confide in is not legally in a different category than an intentional wellness app.

Third, understand the annual reporting requirement. Your legal and product teams need to decide how to count and categorise crisis referrals before January 1, 2027, because the first annual report will cover the first year of operation under the law.

Fourth, review minor-specific design decisions. If any version of your product is accessible to minors, the engagement mechanics need to be assessed against Washington’s standard.

This law joins a rapidly thickening patchwork of state AI regulation. Idaho’s Conversational AI Safety Act is already in effect. Georgia is waiting on gubernatorial signatures for three AI bills with overlapping requirements. The direction is consistent: AI systems interacting with the public are being brought into consumer protection frameworks, state by state, with enforcement tools that do not wait for federal action.

---

If you’re deciding where to start with agents, start here. The free Working With Claude field guide walks through the ecosystem, Claude Code, and a real rollout plan. Get your copy.