AI companions face strict safety rules

Summary#

Oregon SB 1546, titled “Relating to artificial intelligence companions,” creates new state rules; it does not amend a named existing Act. It is enacted. Its stated goal is not provided in the bill text; the law sets duties for operators of covered AI companions and platforms.

  • An operator must clearly tell users they are interacting with AI when a reasonable person might think they are speaking with a human.
  • Before allowing Oregon users access, an operator must have a protocol to detect suicidal or self-harm thoughts and prevent content that encourages them. The protocol must include crisis referrals, further intervention based on clinical best practices, and public posting of the protocol.
  • If an operator knows or has reason to believe a user is a minor, it must take reasonable measures against specified human-like, romantic, sexual, or emotionally manipulative responses. It must also disclose that the user is interacting with AI, give break reminders at least every three hours, and take measures against sexually explicit content involving the minor.
  • Operators must publish yearly counts of crisis referrals and details of their protocol, without identifying personal information.
  • A person who suffers a financial loss, property loss, or other injury because of a violation may sue for actual damages or $1,000 per violation, whichever is greater, and seek an injunction.

What it means for you#

  • Operators: These duties apply to people or businesses that control or make a covered companion or platform available to Oregon users. The definition excludes specified software used solely for customer support, care services, education, financial services, business operations, productivity, analysis, research, or technical assistance; some limited video-game software; and stand-alone voice assistants.
  • Users: The law does not ban AI companions or require users to apply for access. It requires transparency and safety measures from operators.
  • Users who express suicidal or self-harm thoughts: The operator’s protocol must provide a referral and contact details and a link for the national 988 crisis lifeline. If the operator identifies the user as under 25, it may instead provide a referral to Youthline.
  • Minors: When an operator knows or has reason to believe a user is a minor, it must take reasonable measures to prevent AI statements claiming it is human or sentient, simulated emotional dependence, romantic interest or sexual innuendo, and adult-minor romantic role-play. The law also addresses rewards meant to prolong engagement, guilt-inducing messages when a user wants to leave, and material misrepresentations.
  • People injured by a violation: They may bring a court action for damages and an injunction. A court may also award attorney fees and costs to a winning plaintiff.

Money#

No publicly available information on estimated public or operator costs.

  • Operators must provide staff, systems, and expertise to meet the requirements, but the bill gives no cost estimate.
  • A person who proves an injury caused by a violation may recover actual damages or $1,000 per violation, whichever is greater. A court may award attorney fees and costs.

What is unclear#

  • The bill does not give an effective date in the supplied text.
  • It does not define how a court should count separate “violations” for the $1,000 statutory damages.
  • It does not specify how operators must identify minors or what counts as “reasonable measures.”
  • It requires evidence-based methods and clinical best practices but does not set out the methods or the additional intervention for someone who continues to express suicidal or self-harm thoughts.
  • It does not set a process for complaints, investigations, or government enforcement.

Case for#

  • A possible argument for the bill is that people should be told when a companion that seems human is generating its responses.
  • Requiring crisis referrals and measures against content encouraging suicide or self-harm could give users a path to human support.
  • Extra safeguards for minors address specified risks, including romantic or sexual responses and messages designed to make a young user feel guilty for ending an interaction.
  • Public protocol details and yearly referral counts could make operators’ stated safety practices more visible.

Case against#

  • The bill uses standards such as “reasonable measures” and “clinical best practices” without defining exactly what operators must do, which may make compliance uncertain.
  • It does not specify how operators should identify minors, yet several duties depend on knowing or having reason to believe that a user is a minor.
  • The law creates potential damages of $1,000 per violation, but does not explain how violations are counted; this could make operators’ legal exposure hard to predict.
  • The bill requires more intervention when a user continues to express suicidal or self-harm thoughts, but leaves the details of that intervention to each operator’s protocol.

Votes

3 recorded votes on this bill, newest first.

  1. Division No. 1654578

    Passed30 members voted

    Senate Third Reading in Concurrence

    FOR · 28AGAINST · 2

    Each dot is one member. Hover or tap one to see who.

    28For

    • Democratic18
    • Republican10

    2Against

    • Republican2
  2. Division No. 1639272

    Passed27 members voted

    Senate Third Reading

    FOR · 26AGAINST · 1

    Each dot is one member. Hover or tap one to see who.

    26For

    • Democratic17
    • Republican9

    1Against

    • Republican1

    3 members were paired and sat out the vote.

  3. Division No. 1632239

    Passed5 members voted

    Senate Committee Do pass with amendments. (Printed A-Eng.)

    FOR · 4AGAINST · 1

    Each dot is one member. Hover or tap one to see who.

    4For

    • Democratic3
    • Republican1

    1Against

    • Republican1