In the wake of several high-profile lawsuits related to teen suicides allegedly encouraged by AI chatbots, the Senate Judiciary Committee unanimously advanced a bill earlier this month to protect kids from AI-related harms. The GUARD Act, co-sponsored by Sens. Josh Hawley (R.-Mo.) and Richard Blumenthal (D.-Conn.), seeks to safeguard minors by age-gating AI companions designed to simulate human relationships. In addition to advancing young people’s online safety, the bill embodies the best of American federalism, drawing from the strengths of enacted state policies while correcting their weaknesses.
The move by the Senate Judiciary committee reflects a growing popular demand to govern AI in a way that advances human flourishing. Contrary to claims of a growing “patchwork” of hundreds of disparate state AI policies, a new report by the Institute for Family Studies found that AI laws (including on children’s safety) enacted by states between 2023 and 2025 broadly reflect a shared pro-human policy consensus. Hawley and Blumenthal’s bill aligns with this emerging consensus. Although it is novel as a piece of federal legislation, it advances protections enacted by some states while also offering stronger safeguards.
Between 2023 and 2025, eight states passed laws governing AI chatbots in some capacity. At the broadest level, Utah and Maine enacted legislation that requires businesses using customer-facing AI chatbots to disclose the tools’ non-human status. Other laws dealt principally with chatbots marketed for health or therapeutic purposes. For example, Utah, Illinois, and California all enacted laws prohibiting chatbots from being advertised as possessing professional credentials and licenses they do not have. A few states, including Texas, New York, and California, prohibited the promotion of self-harm and suicidal content. Additionally, some states have sought to prevent chatbots from providing or soliciting sexual content from minors.
These laws demonstrate the ability of states to get ahead of emerging issues and indicate strong bipartisan support for safety regulation. Nevertheless, the state laws aimed at protecting children from chatbot harms do not specify the standard of knowledge by which companies knowingly engage with minors. California’s SB 243 (2025) requires AI companies to “institute reasonable measures” to prevent companion chatbots from providing or soliciting sexual content from users “known to be minors.” Likewise, New Hampshire’s HB 143 (2025) specifies that AI companies are liable for “endangering the welfare of a child” if their AI product encourages or solicits sexually explicit conduct, self-harm, illegal drug use, violence, or any other criminal activity. But neither law specifies how AI companies would know that the users are minors.
As the past two decades of internet history have demonstrated, effective liability for tech companies depends on the standards for knowing whether a user is a minor. Currently, under the 1998 federal Children’s Online Privacy Protection Act (COPPA), tech companies are prohibited from collecting data from minors (defined as those under 13) without parental consent. But companies are only deemed to know that a user is a minor if they have “actual knowledge,” which effectively requires them to have viewed a child’s birth certificate or other government documentation in order to be held liable. Although legislation that recently passed the Senate has been touted as updating the 1998 law, it basically maintains the status quo.
To address this loophole, a growing number of states have enacted online safety laws that require age verification or age estimation for pornography, social media, and app stores. While these businesses may be liable on paper, if there is no legally required mechanism for knowing a user is a minor, then liability claims cannot get off the ground. Moreover, because of COPPA’s knowledge standard, social media companies and other websites directed at children have been able to circumvent federal protections for minors.
“The GUARD Act is an example of the benefits of American federalism.”
The GUARD Act carries forward a few significant protections enacted by states for companion chatbots. It requires the disclosure of the AI’s non-human status and non-professional status, prohibits the promotion of physical violence and self harm, and criminalizes the solicitation of sexual content or conduct. Unlike similar state laws, the new bill also solves the knowledge standard issue by requiring that AI companion providers perform age verification. That creates meaningful liability for AI-related harms and thereby obligates compliance by tech companies.
The GUARD Act is an example of the benefits of American federalism, showing that actions at the state and federal level can be mutually reinforcing. Its provisions demonstrate that state policies can helpfully inform federal legislation, but also that federal legislation can address gaps in state policies.