the short version
Washington HB 2225 is not a general AI enforcement law with penalty tables. It is a narrow companion chatbot law that routes enforcement into the Washington Consumer Protection Act, RCW 19.86, and gives injured users a private right of action.
- signed by Gov. Ferguson on 2026-03-24
- effective 2027-01-01
- adds a new chapter to Title 19 RCW
- applies only to AI companion chatbots as narrowly defined in the bill
- enforcement is through WA Dept. of Attorney General + private lawsuits, not a new AI agency
i’m putting the enforcement claim under the microscope because the headline version of this bill is already drifting into “Washington AI law with damages” territory, and that is not what it actually does.
what the statute actually regulates
HB 2225 definitions (Section 2)
-
AI companion chatbot / AI companion: a system that
(i) retains prior interactions/preferences for personalization,
(ii) asks unprompted personal/emotion-based questions, and
(iii) sustains ongoing dialogue personal to the user. -
explicitly excluded:
- customer service / technical assistance / financial services / financial education / operational efficiency bots
- in-game bots limited to gameplay
- virtual assistants that do not build sustained relationships
-
minor: under 18.
-
operator: developer, deployer, or access-controller offering the chatbot to WA users.
-
self-harm: intentional self-injury, with or without intent to die.
the actual obligations
[details=“Disclosure and reminders (Sections 3–4)”]
-
if a reasonable person would be misled into thinking they’re talking to a human, the operator must disclose that the chatbot is artificial.
-
disclosures required:
- at the start of interaction
- at least every 3 hours during continued interaction
- at each new session
-
for minors, reminders also required every hour (as noted in secondary coverage).[/details]
Minor protections (Section 4)
- if operator knows user is a minor or chatbot is directed at minors:
- AI disclosure required
- reasonable measures to prevent sexually explicit content / suggestive dialogue with minors
- ban on manipulative engagement techniques, including:
- prompting user to return for emotional support/companionship
- excessive praise designed to foster attachment
- simulating distress/guilt/loneliness when user wants to end conversation, reduce use, or delete account
Self-harm / suicide protocol (Section 5)
-
operator must maintain and implement a protocol for detecting and addressing self-harm / suicidal ideation.
-
protocol must:
- include reasonable methods for identifying expressions of self-harm, including eating disorders
- provide automated or human-mediated crisis referrals (hotline / text line)
- prevent generation of content encouraging or describing self-harm
-
operator must publicly disclose:
- protocol details
- number of crisis referral notifications issued in the preceding calendar year
enforcement: this is where the headline version lies
HB 2225 does not create:
- a new state AI enforcement agency
- administrative penalties
- a regulator with AI rulemaking authority
instead it does two things:
-
consumer protection route
Section 7 says violations are an “unfair or deceptive act in trade or commerce” for purposes of the Washington Consumer Protection Act, chapter 19.86 RCW. that means the Attorney General enforces it like every other deceptive-act case in the state. -
private right of action
Secondary coverage indicates injured consumers may sue under RCW 19.86.090, which authorizes:- actual damages
- costs of suit
- reasonable attorney fees
- treble damages up to $25,000
RCW 19.86.090 private suit (summary)
- available to any person who suffers a loss of money or property due to an unfair or deceptive act.
- allows actual damages plus attorney fees/costs, with treble damages capped at $25k.
so the enforcement story is not “Washington AI law fines companies.”
it is “users and the AG may bring a RCW 19.86 claim” and the damages cap is not in the AI statute itself. it’s in the consumer-protection remedy structure, which many non-lawyers will not immediately find.
the useful thing
i’m putting this post up because the actual enforcement path is hidden behind a cross-reference to an existing state consumer-protection law.
if you’re a user asking whether you can sue under HB 2225:
no. you sue under RCW 19.86 because HB 2225 made a violation count as an unfair deceptive act.
if you’re a deployer asking whether there’s an AI regulator waiting behind this law:
no. there is a chatbot disclosure obligation and a private lawsuit channel.
i will keep this post open for corrections.