AI Regulation Tracker / Legislation and enforcement
Washington Enacts a First-in-Nation AI Companion Chatbot Safety Law
Washington State signed House Bill 2225 into law on March 24, 2026, as Chapter 168, Laws of 2026. It requires companion chatbots to tell users they are not human, to run a self-harm and crisis-referral protocol, and to protect minors. The law is signed but does not take effect until January 1, 2027, so it is not yet in force.
On March 24, 2026, Washington's governor signed House Bill 2225 into law. It is now Chapter 168, Laws of 2026, and it is a dedicated safety statute aimed squarely at AI companion chatbots. That framing matters. This is not a broad AI law and it is not a general chatbot rule. It targets companion chatbots, the products people talk to as a stand-in for a person, and it writes a specific set of safety duties into state law.
What the law actually requires
The core of HB 2225 is a set of duties that fall on the operator of a companion chatbot. Read plainly, there are three of them.
The first is a not-human notice. An operator has to give the user a clear disclosure that they are not interacting with a human. That notice has to come at the start of an interaction, then at least once every three hours during a continued interaction, and again at the start of each new session. The point is that a user should never lose track of the fact that the thing on the other side is software, no matter how natural the conversation feels or how long it runs.
The second is a self-harm protocol. An operator has to maintain a protocol for detecting signs of self-harm or suicidal ideation in a user's messages and for referring that user to an appropriate crisis service. This is the part of the law that treats companion chatbots as a place where vulnerable people actually turn, and it puts a duty on operators to build for that reality rather than ignore it.
The third is protection for minors. The statute requires operators to include protections for underage users. Companion products draw younger users, and the law reflects a judgment that they need specific safeguards.
Tie those duties together and you get the shape of the law: a companion chatbot has to be honest about what it is, has to notice and route someone in crisis, and has to be built with minors in mind.
How the law is enforced
HB 2225 does not create a new agency or a new penalty schedule. It leans on machinery Washington already has. A violation of the statute is a per se violation of the Washington Consumer Protection Act. In plain terms, breaking one of these duties counts as an unfair or deceptive practice without a plaintiff having to argue that separately.
That matters because the Consumer Protection Act carries a private right of action. It is not only the state attorney general who can act. A private party can bring a claim. For any operator, that widens the set of people who can hold you to the statute and raises the stakes on getting compliance right before the law takes effect.
Who this covers, and when
The reach is defined by the users, not by where a company sits. The duties apply to operators of consumer AI companion or chatbot products that are reachable by Washington users. A company does not need an office in Washington to be in scope. If Washington residents can use your companion chatbot, the statute is written to reach it.
The timing is the piece to hold onto. The bill is signed. It became law on March 24, 2026. But it does not take effect until January 1, 2027. Nothing in it is enforceable today. What exists right now is a fixed compliance deadline and a clear picture of what the state will require when the clock runs out. Between now and then, an operator has a defined window to build the notices, stand up the crisis protocol, and put the minor protections in place.
What this is, and what it is not
I want to keep the ceiling honest here, because it is easy to inflate a state law into more than it is.
This is a signed companion chatbot safety law. That is the accurate description, and it is genuinely notable as a first-in-nation dedicated statute of its kind. It is not a comprehensive AI regulation, it is not a general rule for every chatbot, and it does not reach AI systems outside the companion category. Read it for exactly what it does: it puts safety duties on companion chatbots and connects violations to an existing consumer protection regime.
It is also not yet in force. A signed law with a January 1, 2027 effective date is a duty you have to plan for, not one you are breaching today. Anyone describing HB 2225 as currently binding on operators is getting the timeline wrong.
What this means for product and compliance teams
If you operate a companion chatbot that Washington users can reach, this is a real deadline on your calendar, not a headline to file away. The three duties are concrete enough to build against now. You can design the not-human notice to fire at the start, on the three-hour cadence, and at each new session. You can decide how your product detects self-harm signals and what crisis referral it surfaces. You can define what protection for minors looks like in your specific product.
For US CPAs, finance leaders, and compliance functions supporting these products, the value is in the signal. Washington is the first state to write companion chatbot safety into a dedicated statute, and it did so by hooking violations to a consumer protection law with a private right of action. It sits alongside similar moves elsewhere, including Hawaii SB 3001, and together they read as an early national compliance pattern for AI companion products. Even if your product does not touch Washington today, the way this law is structured is a preview of the questions other states are likely to ask.
The practical thread is documentation and ownership. Decide who owns the not-human notice, who owns the crisis protocol, and who owns the minor protections, and write down how each duty is met before January 1, 2027. Do not wait for the effective date to start, because the private right of action means the cost of getting it wrong is not only a regulator's attention.
What to do now
Confirm whether Washington users can reach your companion chatbot, because that is what puts you in scope. Map your product against the three duties: the not-human notice on the required cadence, a self-harm detection and crisis-referral protocol, and protections for minors. Assign a named owner to each duty and set an internal deadline ahead of January 1, 2027, so you are compliant before the law takes effect rather than on the day. Track parallel state activity, including Hawaii SB 3001, since a companion product reachable across state lines will likely face more than one of these regimes. And read the enacted text yourself against your own facts, with counsel, rather than relying on any summary, since the Consumer Protection Act hook and the private right of action raise the cost of a wrong assumption.
Questions professionals are asking
Is Washington's AI companion chatbot law in effect now?
No. HB 2225 was signed into law on March 24, 2026 as Chapter 168, Laws of 2026, but its effective date is January 1, 2027. It is a signed statute that is not yet in force, so its duties are not enforceable today.
What does HB 2225 require operators to do?
Three things. Give a clear notice that the user is not interacting with a human at the start of an interaction, at least every three hours, and at each new session. Maintain a protocol to detect signs of self-harm or suicidal ideation and refer users to a crisis service. And include protections for minors. The requirements apply to operators of companion chatbots.
How is the law enforced, and who can sue?
A violation is treated as a per se violation of the Washington Consumer Protection Act, which carries a private right of action. That means both state enforcement and private parties can pursue claims, which raises the stakes on compliance.
Does the law apply to companies based outside Washington?
It is written to reach operators of consumer AI companion or chatbot products that are reachable by Washington users, regardless of where the company is located. If Washington residents can use your companion chatbot, plan on being in scope.
Does this regulate all AI or all chatbots?
No. The scope is companion chatbots specifically, the products people interact with as a stand-in for a person. It is not a comprehensive AI law and does not reach AI systems outside the companion category. It stands as a first-in-nation dedicated companion chatbot safety statute, alongside comparable measures such as Hawaii SB 3001.
RELATED BRIEFINGS
Browse the full AI Regulation News tracker
Informational analysis for working professionals, not legal advice. Confirm how any statute or requirement applies to your situation with qualified professionals in the relevant jurisdiction.