Oregon’s new AI companion law makes one thing harder to deny: this is no longer a two-state experiment.
California already enacted a companion chatbot law. New York’s companion safeguards are now in effect. Oregon now adds a third enacted state model through SB 1546, chaptered as Chapter 85.
The three states did not copy one another line for line. They did land on the same basic instinct. When a chatbot is human-like enough that a reasonable person might think they are dealing with a natural person, lawmakers increasingly want disclosure, safety rules, and an enforcement path.
That is why Oregon matters. It makes the pattern easier to see.
For broader companion-chatbot coverage, see Clearon’s earlier piece on AI Companion Safety Laws Are Becoming a Real Compliance Category.
What Oregon Did
Oregon’s SB 1546 is now chaptered as Chapter 85.
The official state materials indicate that the law requires notice when a reasonable person would believe they are interacting with a natural person. The same materials also indicate that a user who suffers ascertainable harm can seek damages and injunctive relief.
That combination matters.
Much of AI regulation is still stuck at the level of agency guidance, draft rules, or future obligations. Oregon’s law is not. It is a state statute aimed at a narrow product category with both a disclosure concept and a private enforcement hook. That makes it a real compliance signal, not just a policy talking point.
The Law Starts At The Relationship Layer
The interesting part is where Oregon starts.
It does not begin with frontier-model debates, general AI risk theory, or a broad licensing framework. It starts with the user experience. If the product is human-like enough that a reasonable person could take it for a natural person, the law cares.
That move is starting to repeat.
California’s companion chatbot law also uses a nonhuman-disclosure model, with additional minors and self-harm safeguards. New York’s law requires conspicuous recurring notices that users are interacting with AI, not a human, along with crisis-intervention protocols. Oregon now reinforces the same basic idea from another direction: do not let a relationship-style AI system pass as human without legal consequences.
This is why companion-chatbot regulation looks different from a lot of other AI law. The pressure point is not only model capability. It is simulated human interaction.
Why Oregon Matters Beyond Oregon
One enacted state law can be dismissed as an outlier. Three enacted state models are harder to wave away.
That does not mean every state will use the same definitions or remedies. It means companies now have more reason to assume that relationship-like AI systems will keep drawing targeted legislation, especially where minors, self-harm, dependency, sexual content, or emotionally manipulative design are in the frame.
Oregon also helps confirm that nonhuman notice is becoming the floor.
The disclosure duty is the first thing lawmakers can agree on. If the system feels human, tell the user it is not. After that, the next questions usually follow fast:
- what happens when a user shows signs of crisis;
- what the product does for minors;
- whether the design rewards emotional dependence;
- what marketing claims were made about safety or support; and
- what records the company can produce if a regulator or plaintiff asks how the product was reviewed.
That broader pattern already appears in Clearon’s recent FTC companion-chatbot coverage and in the other state companion laws. Oregon fits squarely inside it.
The Private Enforcement Angle Matters
The Oregon measure overview’s reference to damages and injunctive relief is one reason this story deserves its own article.
Disclosure rules matter on their own. Disclosure rules backed by a harmed-user action create a sharper litigation question. They raise the stakes for product teams that still treat chatbot identity notices as soft UX copy rather than compliance language tied to a product design theory.
That does not mean every case will be easy to prove. It does mean the compliance conversation changes when a statute gives users an avenue to claim harm from noncompliance.
For in-house teams, that makes Oregon more than a notice law. It is a warning that chatbot identity, safety design, and consumer expectations can become plaintiff-side issues as well as regulatory ones.
What Companies Should Review Now
Companies offering companion or emotionally responsive chatbots should not treat Oregon as a one-off state update to file away.
They should use it as a trigger to review:
- whether any product could reasonably be perceived as a natural person or relationship-style companion;
- where nonhuman notices appear, how clear they are, and whether they recur when interactions continue;
- how the product handles minors, emotionally vulnerable users, and crisis scenarios;
- whether marketing or onboarding language overstates safety, support, or human-like qualities;
- whether engagement design could be framed as encouraging dependence or extended emotional reliance; and
- what internal records exist showing how those decisions were made before launch.
The records point matters. A lot of companion-AI risk is turning into a proof problem. If a company says it disclosed the system’s nature, tested foreseeable harms, and built safeguards, it should be able to show the file.
That is also where Oregon connects to the FTC’s 6(b) inquiry. The law and the inquiry use different tools, but they are pushing toward the same practical result: companies should expect to explain how relationship-like AI products were designed, disclosed, tested, and governed.
Bottom Line
Oregon’s new AI companion law is not just another state headline.
It is more evidence that chatbot regulation is moving to the relationship layer. If a system is built to feel human, keep users engaged, and occupy an emotionally salient role, lawmakers are increasingly treating that as its own legal problem.
For companies building companion-style AI, the lesson is direct. A generic chatbot disclosure and a moderation policy are not enough. Oregon suggests that states are looking for something more specific: clear nonhuman notice, a defensible safety approach, and an enforcement path when those basics fail.