AI chatbots must disclose they're not human

The Oregon Legislature approved file SB1546 ("AI chatbots must disclose they're not human"), by a recorded vote of 80 to 2. Last action: April 6, 2026.

Oregon file SB1546, before the Oregon Legislature. Passed as of 2026-04-06.

What it does

You'll see a clear notice that you're talking to AI When using an AI chatbot or companion app, operators must disclose in plain terms that you're interacting with software, not a person, if a reasonable user would otherwise think they were talking to a human.

The bigger picture

AI platforms must screen and intervene on suicide risk Operators must implement protocols to detect suicidal or self-harm ideation, refuse harmful outputs, and provide crisis lifeline referrals—shifting responsibility for user mental health safety onto AI companies rather than users alone.

What to watch

Enforcement relies on lawsuits; compliance costs unclear Users can sue for damages if harmed, but the law's definition of 'reasonable person' for disclosure and 'ascertainable harm' may lead to contested litigation, and implementation costs for detection protocols are not specified.

Official summary

Digest: Tells those who make AI software to tell users that the users are talking to software, not a human. Tells them they must try to prevent users from getting output that causes suicidal feelings or thoughts. (Flesch Readability Score: 71.0). Requires operators of artificial intelligence companions and artificial intelligence companion platforms to provide notice to users that the users are interacting with artificial output if a reasonable person that interacts with the artificial intelligence companion or artificial intelligence companion platform would believe that the person was interacting with a natural person. Requires the operators to have in place a protocol for detecting suicidal ideation or intent or self-harm ideation or intent and to prevent output that could cause such ideation or intent in users. Specifies minimum contents of the protocol<b>, including referral to an appropriate crisis lifeline and additional intervention informed by clinical best practices and expertise</b>. Requires an operator to make certain statements and disclosures if the operator has reason to believe that a user that interacts with the operator's artificial intelligence companion or a

How they voted

Yes (80)

No (2)

Did not vote / other (8)

Read the official record for SB1546

See everything up for a vote · This week in Oregon · See what we have filed with the clerk