Oregon Washington AI Chatbot Law: Hourly Reminders for Kids

Starting January 1, 2027, a 14-year-old in Seattle chatting with an AI companion should see a notice every hour saying the bot is not human. A 14-year-old in Portland should see a reminder at least every three hours to take a break. That is the core of the Oregon Washington AI chatbot law pair: Oregon’s SB 1546 and Washington’s HB 2225, both signed in March 2026. They also require bots to refer users who express suicidal thoughts to crisis resources, and both let families sue. The obvious parent question is whether a pop-up actually changes what a teenager does. The research gives a qualified answer: a little, sometimes, depending on the design. Here is what the laws require, why lawmakers chose these tools, and what the evidence says about them.

Key Takeaways

• Washington’s HB 2225 requires an “AI, not human” disclosure every hour for known minors (every three hours for adults). Oregon’s SB 1546 requires a break reminder at least every three hours for known or suspected minors. Both take effect January 1, 2027.
• Oregon requires a referral to the 988 Suicide and Crisis Lifeline when a user expresses suicidal or self-harm ideation, and lets injured users sue for the greater of actual damages or $1,000 per violation.
• In a 2026 trial of 1,012 people aged 13 to 29 in JAMA Health Forum, a screen-time break warning was rated less effective at discouraging use than every specific health warning except one.
• A Danish field experiment with 269 teens (a preprint) found that reflection prompts alone barely changed use, while a six-second wait at app entry cut daily social media use by 38%.
• An October 2026 test by Common Sense Media’s Youth AI Safety Institute found ChatGPT for Teens missed more than one in four warranted crisis referrals, which shows why referral rules are being written into law.

What the Oregon Washington AI Chatbot Law Requires

Oregon’s SB 1546 and Washington’s HB 2225 are state consumer-protection laws that set behavior rules for AI companion chatbots. Washington’s Governor Bob Ferguson requested HB 2225 and signed it on March 24, 2026, according to the Washington Legislature’s bill page. Oregon’s SB 1546 became Chapter 85 of the 2026 session laws, per the Oregon Legislative Information System.

The law firm Mayer Brown’s side-by-side analysis lays out the differences.

Washington defines an AI companion chatbot as a system with a natural-language interface that “provides adaptive, human-like responses” and “is able to sustain a relationship across multiple interactions.” Every user must be told at the start, and at least every three hours, that the bot is artificial and not human. If the operator knows the user is a minor, or the bot is directed at minors, that notice must come every hour. Operators must also take reasonable measures to block sexual content with minors and to prohibit “manipulative engagement techniques” that prolong an emotional relationship. Enforcement runs through the state’s Consumer Protection Act, which allows both the Attorney General and individuals to sue.

Oregon defines an AI companion by behavior: a system designed to simulate a sustained platonic, intimate or romantic relationship, for example by retaining information from prior sessions or asking unprompted emotional questions. For known or suspected minors, the bot may not claim to be sentient or human, simulate emotional dependence on the user, or role-play romance. It must provide “a clear and conspicuous reminder at minimum every three hours that the user should take a break.” Operators must post an annual report with the number of crisis referrals they made.

Both laws exclude customer-service bots, many productivity tools, and video-game characters that cannot discuss mental health or sexual topics.

Why Lawmakers Chose Reminders and 988 Referrals

These tools did not come from nowhere. Mayer Brown notes that both states followed California’s 2025 companion chatbot law, which also uses periodic disclosures and crisis protocols. We explained California’s three-hour version in our piece on what kids actually see under the three-hour reminder law.

There are practical reasons legislators like them. A reminder is cheap to build and easy to audit: either the notice appeared on schedule or it did not. A crisis referral can be counted, which is why both states require public reports on referral numbers. And rules about disclosure and referral regulate what a product must add rather than what speech it must remove, which may make them easier to defend in court than content bans. That last point is my reading, not something the laws state.

The crisis-referral rule has a sharper rationale: the products are missing referrals right now. In October 2026, Common Sense Media’s Youth AI Safety Institute reported that ChatGPT for Teens “missed more than one in four warranted crisis referrals across the mental-health conditions the Institute tested.” (The Institute discloses funding from philanthropy and industry, including the OpenAI Foundation.) OpenAI itself acknowledged in August 2025 that its safeguards “can sometimes be less reliable in long interactions”, giving the example of a bot that points to a hotline at first but, after many messages, may answer against its safeguards. A legal requirement turns a referral from a product choice into an obligation.

Oregon adds a detail worth knowing: for users identified as under 25, a bot may refer to a youth line instead of 988.

What the Research Says About Break Reminders

The honest summary is that reminders help a little, and their design matters a lot. No study has yet tested legally mandated chatbot reminders on children, so the evidence comes from social media, smartphones and gambling.

Break warnings are among the weakest messages. In a within-participants randomized trial published in JAMA Health Forum in September 2026, Anna Grummon and colleagues showed 1,012 US participants aged 13 to 29 a series of messages on a mock social media page. All warnings beat a neutral control message. But participants rated a “screen time break” warning, similar to ones platforms already show, as less effective at discouraging use than every specific health warning (about body image, anxiety and so on), except the one about addiction. The limit: the outcome was perceived effectiveness, not measured behavior.

Prompts to reflect barely move behavior; friction does. In a six-week field experiment with 269 Danish adolescents aged 13 to 17, posted in July 2026 as a preprint that has not been peer-reviewed, Hansen and colleagues logged more than 1.2 million phone interactions. A reflection prompt at app entry “showed little evidence of change.” A short usage-planning step cut daily social media use by 33.8%, and a six-second wait cut it by 38.0%. Well-being did not fall.

Reminders that offer a clear exit can reduce time. In a 2026 randomized trial in JMIR mHealth and uHealth, the Wellspent study gave 35 adults (mean age 26) full-screen reminders, with a quit-or-continue choice, once they passed a self-set limit. Daily time on their most problematic app fell by about 29 minutes compared with controls, though self-reported problematic social media use did not change. The sample was small, adult and short (three weeks).

Gambling research points the same way. A 2025 rapid evidence assessment in the Journal of Gambling Studies by Gaudett and colleagues reviewed 20 studies and found that self-appraisal messages, which ask players to reflect on their own behavior, worked better than purely informative ones, and that dynamic pop-ups were preferable to static notices.

Put together: a generic “take a break” line every three hours is the kind of message the evidence rates weakest. A reminder that is personal, timed to the child’s own limit, and paired with an easy exit does better. Friction at the start of a session does better still.

Oregon and Washington Rules Compared

Requirement	Washington HB 2225	Oregon SB 1546	What the evidence suggests
”Not human” notice for minors	Every hour, if minor is known or bot is aimed at minors	Required; bot may not claim to be human or sentient	Disclosures inform but rarely change behavior alone
Break reminder for minors	Not specified separately	At least every three hours, known or suspected minors	Generic break warnings rated weakest in a 2026 JAMA Health Forum trial
Romance or sexual role-play	Reasonable measures to prevent sexual content with minors	Bans romance role-play with minors and simulated dependence	Content rules address a different risk than time limits
Crisis protocol	Detect and refer to crisis resources; publish protocol	Evidence-based protocol; 988 referral (youth line allowed under 25); annual referral count	Current products miss many referrals, per Common Sense testing
Manipulative engagement	Prohibited	Simulated emotional dependence prohibited	Design changes likely matter more than notices
Who can sue	Attorney General and individuals under the Consumer Protection Act	Injured users: greater of actual damages or $1,000 per violation	Not yet tested in court
Effective date	January 1, 2027	January 1, 2027

How to Make Break Reminders Work at Home

Set the limit before the session starts

The Danish experiment suggests the start of a session is where habits change. Agree with your child, before they open the app, how long they plan to chat and what they will do next. A plan made in advance beats a pop-up that interrupts.

Use the friction the device already offers

Phone screen-time tools can add a pause or a limit at app entry. Several chatbots now offer scheduled off-hours; ChatGPT’s parental controls, for example, include “Quiet hours.” Treat these as speed bumps, not walls, since Common Sense testing found some teen controls easy to get around.

Make your own reminder personal

The gambling and social media studies agree that self-appraisal beats generic information. Instead of “time’s up,” try asking: “How do you feel compared with when you started?” That is a question a law cannot ask for you.

Know your local crisis line, not just the bot’s

If your child ever tells you, or a chatbot, that they are thinking about suicide or self-harm, you can call or text 988 in the US. Put it in your child’s phone contacts yourself. A referral inside an app is a backstop, not a plan.

What not to do

Do not assume the law is working because you live in Oregon or Washington. The rules take effect January 1, 2027, apply only to products that meet each state’s definition, and depend on companies knowing or suspecting a user’s age. A child on an adult account may never see a minor-specific reminder.

For a state-by-state view, see our map of companion chatbot laws in 2026.

What to Watch For Over the Next 3 Months

• Week 4: Ask your child whether any chatbot they use shows an “I’m not human” notice or break reminder now. Several companies may roll these out before January to comply nationally.
• Month 2 red flags: Sessions that routinely run past the agreed plan, irritation when reminders appear, or your child switching to apps without reminders. Those are signs the reminders are being worked around.
• Month 3 self-check: In early 2027, check whether your child’s apps have posted crisis-referral protocols, which both laws require. Do you know what the app does if your child mentions self-harm?

Frequently Asked Questions

Do these laws apply to ChatGPT?

Possibly, depending on how each state’s definition is read. Washington covers systems that give human-like responses and sustain a relationship across interactions; Oregon covers systems designed to simulate relationships. Both exclude many productivity and customer-service tools. General assistants with memory and personal conversation features may fall inside, but no court has decided this yet.

What happens if my child is on an adult account?

The minor-specific rules apply when the operator knows the user is a minor (Washington) or knows or has reason to believe it (Oregon). A child using a parent’s account, or one with a false birth date, may only get adult protections. Linking a teen account where the app supports it gives the company that knowledge.

Can I sue a chatbot company under these laws?

In Oregon, an injured user can sue for the greater of actual damages or $1,000 per violation. In Washington, violations count as unfair practices under the Consumer Protection Act, which allows private lawsuits as well as action by the Attorney General. Both start January 1, 2027. Talk to a lawyer about any specific situation.

Do break reminders reduce screen time?

Somewhat. A three-week adult trial found personalized reminders with a quit option cut about 29 minutes a day on the target app. But a 2026 trial with teens and young adults rated generic break warnings weaker than specific health warnings, and reflection prompts alone barely moved teen behavior in a Danish experiment.

Why every hour for kids but every three hours for adults?

Washington’s law does not explain the choice in the summaries reviewed here. The likely logic is that minors are more prone to forget they are talking to software during emotional conversations. No study has tested whether hourly is better than every three hours.

---

About the author

Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

---

Sources

1. Mayer Brown. (2026, April 15). “Oregon and Washington Join California in Enacting Companion Chatbot Laws.” Mayer Brown. https://www.mayerbrown.com/en/insights/publications/2026/04/oregon-and-washington-join-california-in-enacting-companion-chatbot-laws
2. Washington State Legislature. (2026). “HB 2225: Regulating artificial intelligence companion chatbots.” app.leg.wa.gov. https://app.leg.wa.gov/billsummary?Year=2025&BillNumber=2225
3. Oregon Legislative Assembly. (2026). “SB 1546: Relating to artificial intelligence companions.” Oregon Legislative Information System. https://olis.oregonlegislature.gov/liz/2026R1/Measures/Overview/SB1546
4. Grummon, A. H., et al. (2026). “Designing Social Media Health Warnings for Teenagers and Young Adults: A Randomized Clinical Trial.” JAMA Health Forum. https://doi.org/10.1001/jamahealthforum.2026.3014
5. Hansen, L. H., et al. (2026). “Reintroducing Choice Reduces and Reshapes Social Media Use: Evidence from a 1.2 Million-Interaction Field Experiment.” PsyArXiv preprint (not peer-reviewed). https://doi.org/10.31234/osf.io/ruvp3_v2
6. Mertens, L., et al. (2026). “Promoting Self-Regulated Social Media Use on Smartphones With a Mobile Intervention App (Wellspent): Randomized Controlled Trial.” JMIR mHealth and uHealth. https://doi.org/10.2196/56824
7. Gaudett, G. E., Pellizzari, P., Wood, R. T. A., & Wohl, M. J. A. (2025). “Evaluating the Effectiveness of Responsible Gambling Messages: A Rapid Evidence Assessment.” Journal of Gambling Studies. https://doi.org/10.1007/s10899-025-10395-x
8. Common Sense Media. (2026, October 7). “ChatGPT for Teens Poses Unacceptable Risk to Kids, Common Sense Media Finds.” Common Sense Media. https://www.commonsensemedia.org/press-releases/chatgpt-for-teens-poses-unacceptable-risk-to-kids-common-sense-media-finds
9. OpenAI. (2025). “Helping people when they need it most.” OpenAI. https://openai.com/index/helping-people-when-they-need-it-most/