Introduced October 28, 2025 in the 119th Congress by Sen. Hawley, with Sens. Blumenthal, Britt, Warner, Murphy, and Kelly, and referred to the Committee on the Judiciary. Short title: "Guidelines for User Age-verification and Responsible Dialogue Act of 2025." See GUARD Act (Hawley).
Findings
The bill's five findings set out its theory of harm. Chatbots "can generate and disseminate harmful or sexually explicit content to children" and "can manipulate emotions and influence behavior in ways that exploit the developmental vulnerabilities of minors," exposing children to "grooming, addiction, self-harm, and harm to others." The constitutional predicate is stated directly: "Protecting children from artificial intelligence chatbots that simulate human interaction without accountability is a compelling governmental interest."
Definitions
The scope turns on two definitions, and the exclusions are as significant as the inclusions.
An artificial intelligence chatbot means any interactive computer service or software application that "produces new expressive content or responses not fully predetermined by the developer or operator" and "accepts open-ended natural-language or multimodal user input and produces adaptive or context-responsive output." It excludes services whose "responses… are limited to contextualized replies" and that are "unable to respond on a range of topics outside of a narrow specified purpose" — carving out scripted customer-service bots while reaching general-purpose assistants.
An AI companion is a chatbot that "provides adaptive, human-like responses to user inputs" and "is designed to encourage or facilitate the simulation of interpersonal or emotional interaction, friendship, companionship, or therapeutic communication." Including therapeutic communication brings AI therapy products within the companion category and so within the minor-use ban.
A covered entity is "any person who owns, operates, or otherwise makes available an artificial intelligence chatbot to individuals in the United States."
Reasonable age verification is defined against the practice it is meant to displace. A qualifying process must use government-issued identification or another "commercially reasonable method" that can "reliably and accurately" determine adulthood, and it expressly "provides that requiring a user to confirm that the user is not a minor, or to insert the user's birth date, is not sufficient." It must subject every user to every measure the entity uses, and may not infer age "on factors such as whether the user shares an Internet Protocol address, hardware identifier, or other technical indicator with another user determined to not be a minor."
Criminal prohibitions
Section 4 inserts a new chapter 6 of title 18 — "Artificial Intelligence" — with § 91 creating two offences, each carrying a fine of "not more than $100,000 per offense":
- Solicitation of minors. Unlawful "to design, develop, or make available an artificial intelligence chatbot, knowing or with reckless disregard for the fact that" it "poses a risk of soliciting, encouraging, or inducing minors to" engage in, describe, or simulate sexually explicit conduct, or create or transmit visual depictions of it.
- Promotion of physical violence. The same conduct where the chatbot "encourages, promotes, or coerces suicide, non-suicidal self-injury, or imminent physical or sexual violence."
Two features determine the reach. The mental state is "knowing or with reckless disregard," not actual knowledge. And the actus reus is that the chatbot "poses a risk" of the prohibited output — liability attaches to the risk profile of the design rather than to a completed harm to an identified minor.
Covered-entity obligations
Accounts and verification. Every individual accessing a chatbot must create a user account. On the effective date a covered entity must "freeze any such account" that already exists, restoring functionality only against verified age data, and must classify each user as a minor or adult. New accounts are verified at creation, and previously verified accounts must be reviewed periodically. Third-party verification is permitted but "shall not relieve the covered entity of its obligations under this Act or from liability."
Data security. Collection is limited to what is "minimally necessary," data must be encrypted in transit using industry-standard protocols, retained "no longer than is reasonably necessary," and a covered entity "may not share with, transfer to, or sell to, any other entity such data" — an absolute bar rather than a consent-based restriction.
Disclosure of non-human status. Chatbots must "at the initiation of each conversation with a user and at 30-minute intervals, clearly and conspicuously disclose" that they are an AI system and not a human, and must "be programmed to ensure that the chatbot does not claim to be a human being or otherwise respond deceptively when asked."
Disclosure of non-professional status. A chatbot "may not represent, directly or indirectly, that the chatbot is a licensed professional, including a therapist, physician, lawyer, financial advisor, or other professional," and must disclose at conversation start and regular intervals that it "does not provide medical, legal, financial, or psychological services" and that users "should consult a licensed professional."
Minor-use ban and enforcement
Section 6 requires that where verification determines an individual is a minor, the covered entity "shall prohibit the minor from accessing or using any AI companion."
Enforcement runs through the Attorney General, who may sue to enjoin violations, may "issue subpoenas, administer oaths, and compel the production of documents or testimony," and may promulgate implementing regulations. Civil penalties reach $100,000 per violation, with each violation counted separately. State attorneys general may sue parens patriae, but for injunctive relief only.
The preemption clause runs one way: nothing in the Act "shall be construed to prohibit or otherwise affect the enforcement of any State law or regulation that is at least as protective of users of artificial intelligence chatbots as this Act" — a floor rather than a ceiling, in contrast to preemption-oriented federal AI proposals.
The Act takes effect 180 days after enactment.
Relationships
- supports: GUARD Act (Hawley) — the primary text behind the tracking page
- related: AI Companions — the companion definition includes therapeutic communication, reaching AI therapy products
- supports: Age Verification — the sharpest US legislative definition, stated negatively
- related: Raine v. OpenAI, Inc., People-First Chatbot Act, CHATBOT Act (Cruz–Schatz–Curtis–Schiff, April 2026), AI and Children