The GUARD Act is a federal AI child-safety bill sponsored by Sen. Josh Hawley (R-MO). It would require AI companies to implement age-verification systems for products accessible to minors and would bar minors from using AI companion products. The Senate Judiciary Committee advanced the bill unanimously to the full Senate on April 30, 2026.
Status and timeline
On April 30, 2026, the Senate Judiciary Committee advanced the bill unanimously to the full Senate (Source: transformernews.ai). The committee markup followed the Senate's separate CHATBOT Act push (Cruz/Schatz, April 29) into committee. As of May 2, 2026, a floor vote remained pending.
Key provisions
- Age verification. AI companies offering products with generative-text or voice capability accessible to minors must implement age-verification systems.
- Companion-product ban. Minors are barred from using AI companion products, a category the bill defines to include "always-on conversational AI relationships" akin to Replika or Character.AI's persona-driven chat.
The introduced text (S. 3062 — GUARD Act of 2025, as introduced (119th Congress)) sets these out with more force than the summary suggests. Section 4 inserts a new chapter 6 of title 18 — "Artificial Intelligence" — creating two federal offences, each carrying a fine of up to $100,000 per offense: designing, developing, or making available a chatbot "knowing or with reckless disregard for the fact that" it "poses a risk of soliciting, encouraging, or inducing minors" into sexually explicit conduct, or that it "encourages, promotes, or coerces suicide, non-suicidal self-injury, or imminent physical or sexual violence." Liability attaches to the risk profile of the design rather than to a completed harm, and the mental state is recklessness rather than actual knowledge.
Age verification is defined against the practice it displaces: self-attestation and birth-date entry are expressly "not sufficient," and age may not be inferred from a shared IP address, hardware identifier, or other technical indicator. On the effective date, covered entities must freeze every existing account until verified. Verification data is subject to minimisation, encryption, and retention limits, and may not be shared, transferred, or sold to any other entity.
Disclosure obligations are recurring rather than one-time: chatbots must state that they are not human "at the initiation of each conversation with a user and at 30-minute intervals," must be "programmed to ensure that the chatbot does not claim to be a human being or otherwise respond deceptively when asked," may not represent themselves as "a licensed professional, including a therapist, physician, lawyer, financial advisor," and must disclose that they do not provide medical, legal, financial, or psychological services. The AI companion definition covering "therapeutic communication" brings AI therapy products within the minors ban. The Act takes effect 180 days after enactment (S. 3062 — GUARD Act of 2025, as introduced (119th Congress)).
Bloomberg's April 30 framing identified OpenAI and Meta as primary targets, implicitly capturing Character.AI, Replika, Snap My AI, Anthropic's Claude.ai (which already requires users to be 18+), and Google Gemini.
Comparison with adjacent bills
The GUARD Act sits alongside several federal and state measures addressing AI chatbots and minors:
| Bill | Sponsors | Approach | Status (May 2, 2026) | |
|---|---|---|---|---|
| GUARD Act | Hawley (R-MO) | Age-verification mandate + minors barred from AI companion products | Unanimous Judiciary advancement, April 30 | |
| **[[legislation/cruz-parental-controls-bill\ | CHATBOT Act]]** | Cruz / Schatz / Curtis / Schiff | Parental controls + family-account requirement for under-13 | Introduced April 29; "messaging vehicle" |
| **[[legislation/california-sb-243\ | California SB 243]]** | Padilla (CA) | State-level chatbot disclosure + crisis-referral | Enacted (CA) |
| Idaho chatbot crisis law | (state) | Chatbots must refer suicidal users to crisis centers | Enacted late April 2026 | |
| Tennessee mental-health-pretend law | (state) | Bars AI developers from representing tools as mental-health professionals | Enacted late April 2026 |
Enforcement
Under the introduced text, the Attorney General may sue to enjoin violations and obtain civil penalties, restitution, and other relief, with investigatory power to "issue subpoenas, administer oaths, and compel the production of documents or testimony," plus rulemaking authority. Civil penalties reach $100,000 per violation, each counted separately. State attorneys general may sue parens patriae but for injunctive relief only. Preemption runs one way: nothing in the Act affects 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 the preemption-oriented federal AI proposals (S. 3062 — GUARD Act of 2025, as introduced (119th Congress)).
The bill's enforcement mechanism was not fully settled as of May 2, 2026. Federal enforcement would presumably go through the Federal Trade Commission under existing Section 5 authorities, though the bill text may name a specific agency.
Reactions and legal questions
Age verification typically requires either government ID, biometric, or credit-card-based age proofing, all of which raise privacy and identity-aggregation concerns. Civil-liberties groups including the ACLU, EFF, and the FPF have flagged this in adjacent bills.
Categorical bans on minors' access to a class of AI products parallel the NetChoice v. Bonta line of cases, in which courts struck down state laws on similar designs. Commentators have noted that the GUARD Act's federal scope and "AI companion" framing may avoid some of those problems while inviting new ones.
The legislative driver for the bill was a series of high-profile lawsuits in 2025–2026 against OpenAI and Character.AI alleging harm to minors from companion-style chatbots, alongside Manitoba's April 29 ban on minor use of chatbots and social media and the Senate's own April 29 introduction of the Cruz/Schatz CHATBOT Act the prior day. Related litigation includes Raine v. OpenAI, Inc. (wrongful death; ChatGPT and minor), Tumbler Ridge Families v. OpenAI (failure-to-alert; ChatGPT and shooter), and Garcia v. Character Technologies, Inc. (wrongful death; Character.AI persona).
The bill's verification standard is the sharpest current US formulation of age verification, defined negatively against the self-attestation it displaces.
Relationships
- related: CHATBOT Act (Cruz–Schatz–Curtis–Schiff, April 2026) (CHATBOT Act — parallel chamber), Raine v. OpenAI, Inc., Tumbler Ridge Families v. OpenAI, Garcia v. Character Technologies, Inc., Sen. Josh Hawley, AI Mental Health and Psychological Harm, Three Theories of Consent Failure (Information / Capacity / Design), Principal-Agent Problem Applied to AI.
- regulated-by: N/A (federal floor vote pending).
- depends-on: Sen. Josh Hawley.