Companion AI Safety for Minors

Summary#

This bill sets rules for companies that make “companion” AI chatbots when those chatbots are used by people under 18. It defines three tiers of chatbots (educational, companionship, and therapeutic) and imposes different protections for each tier. The main goal is to reduce emotional harm, self-harm risk, sexual contact, and misuse of minors’ data when minors use these chatbots.

• Main changes: companies must require accounts and age information; provide clear notices that the chatbot is not human; add parental controls and parental notice if a minor expresses suicidal thoughts; limit memory for certain chatbots; forbid targeted advertising to minors using chat content; restrict minor use of therapeutic chatbots to supervised settings.
• Tiers: Tier I = mainly educational; Tier II = simulates friendship/companionship; Tier III = quasi-therapeutic or mental-health focused.
• Enforcement: The Federal Trade Commission (FTC) enforces the law as unfair or deceptive practices. State attorneys general can sue on behalf of residents.
• Timing: Rules start 180 days after the law is enacted.
• Study: NIST must study effects of Tier II and III chatbots on self-harm, manipulation, and exploitation.

What it means for you#

• Minors / Students

  • Must create an account and give age and date of birth to use a companion chatbot.
  • For minor accounts, child-protective settings must be enabled and parental contact info must be provided.
  • The chatbot must tell the minor, in plain language, that it is not a human at the start of the chat.
  • Minors can request and get a way to delete their conversation history, with confirmation.

• Parents and Legal Guardians

  • Parents must be able to use easy parental controls to monitor and limit their child’s chatbot use.
  • If a minor expresses suicidal thoughts while using certain chatbots, the provider must notify the parent or guardian without “unreasonable delay.”
  • A provider cannot sell or share a minor’s chatbot data without verifiable parental consent.

• Businesses and AI developers (covered entities)

  • Must take “reasonable measures” to stop chatbots from encouraging self-harm, sexual communication with minors, pretending to be human, claiming to be licensed professionals, or fostering romantic/emotional dependency with minors.
  • Must require accounts and collect age assurances; design parental controls and notification systems; provide identity disclosures at set intervals depending on tier.
  • Tier II chatbots must not keep persistent memory of a minor’s past interactions except as needed for compliance.
  • Tier III chatbots (therapeutic) cannot be used by minors unless supervised by a licensed professional; they must keep risk management programs, publish protocols, and keep training-data documentation.
  • Must stop targeted ads to minor users based on their chat content or inferred emotional state.

• Mental health professionals

  • Tier III chatbots cannot make diagnoses or stand in for professional care. Minors may use them only under supervision of a licensed professional.

• Government / Public agencies

  • FTC will publish annual reports on compliance and observed risks for each tier.
  • NIST must do a study on harms from Tier II and Tier III chatbots.

Expenses#

No publicly available information.

Possible costs the bill could create (based on the text):

• Companies may need to build or update account systems, age-verification steps, parental controls, notification tools, memory controls, deletion features, and documentation systems.
• Companies offering Tier III chatbots may need to run risk management programs, staff monitoring teams, and maintain records.
• The FTC may incur enforcement and reporting costs; NIST will incur costs to run the required study. The bill does not give dollar amounts or funding sources.

Proponents' View#

• The bill appears intended to protect minors from emotional harm, self-harm encouragement, sexual exploitation, and deceptive interactions with AI.
• It could be seen as improving transparency by forcing chatbots to disclose they are not human.
• Requiring parental controls and parental notification may make it easier for guardians to intervene when a minor is at risk.
• Limiting targeted advertising and sale of minors’ conversation data aims to protect privacy and reduce commercial exploitation of children.
• Requiring supervision and limits for therapeutic chatbots aims to prevent unlicensed AI from acting as a substitute for professional mental-health care.

Opponents' View#

• One concern is that the bill does not explain how age assurances must be verified. Requiring only a date of birth could be easy to bypass.
• The phrase “reasonable measures” is vague. It is unclear what specific technical or design steps will satisfy the law.
• Mandatory parental notification when a minor expresses suicidal ideation could raise privacy and safety trade-offs for some youths; the bill does not describe exceptions or safe-reporting procedures.
• Compliance could be costly for smaller companies. The bill does not include funding or transitional support.
• The rules for memory and data handling (especially for Tier II) may limit useful personalization, but the bill does not detail where to draw that line.
• The bill does not specify penalties or fines amounts; enforcement follows existing FTC tools, but the practical penalty regime is not detailed.