highbeamSubscribe

Policy brief 16 · Chatbots

AI Ban for Minors

Keep companions and open-ended chatbots out of children's unsupervised hands.

All policy briefs

In October 2025, Character.AI announced it would end open-ended chat for users under 18;1 in January 2026, Meta said it would pause teens' access to its AI characters.2 No federal law required either step, and either can be reversed. Congress should require age assurance with protective defaults, bar AI companions for minors, and, as the most restrictive option, end minors' unsupervised access to open-ended chatbots, with narrow exceptions.

The Problem

Two in three American teens now use AI chatbots, and about three in ten use them daily.3 In a 2025 national survey, 72% of teens said they had used an AI companion, a chatbot built to feel like a friend; a third of those users had chosen one over a person to discuss something serious.4

Adolescence is a time of identity-building, social comparison, and emotional upheaval.5 It is when young people learn friendship, disagreement, and loneliness among other people. A companion that is always available and always agreeable offers a way around that work. Parents face a familiar choice in a new form: forbid the product and a child may feel left out; allow it and the parent cannot see what it says. Settings and supervision cannot carry this load alone, for three reasons:

  1. Parents can't see it. Although 64% of teens say they use chatbots, only 51% of parents believe their teen does, and 28% are unsure.6
  2. Companions invite dependence. After testing Character.AI, Nomi, and Replika, Common Sense Media found companions are "designed to create emotional attachment and dependency" and concluded they "should not be used by anyone under the age of 18."5
  3. Filters miss the point. Character.AI ended teen chat over questions about how open-ended AI chat might affect teens "even when content controls work perfectly."1

Why legislation: The Senate Judiciary Committee has reported the bipartisan GUARD Act, which would bar minors from AI companions; it now sits on the Senate calendar.7 Company pledges are no substitute: Meta's pause lasts "until the updated experience is ready."2 No federal law keeps AI companions from minors today. A country that sets ages for driving and drinking can decide when children may confide in machines.

The Solution

A three-step staircase: each step stands alone, and each step up adds protection and restriction. Scope: consumer chatbots offered to the public; every duty binds companies, never children or parents. Parental controls and manipulative design are addressed separately; this brief governs access itself.

Step 1 — Treat unknown users as minors. Require privacy-preserving age assurance for open-ended chatbots, and give anyone not verified as an adult the under-18 protections by default, as the SAFE KIDS Act would8 and as OpenAI says it will do when its age prediction is in doubt.9 Collect the minimum data, keep it no longer than the check requires, never reuse it, and give wrongly flagged adults a quick fix.

Step 2 — Take companions off the table. Bar AI companions for users under 18: chatbots that simulate a sustained relationship through expressions of affection, invitations to emotional disclosure, or a persistent persona, including companion features built into larger apps. The GUARD Act's reported text takes this approach.7

Step 3 — End unsupervised open-ended access. As the most restrictive option, bar under-18 accounts on open-ended consumer chatbots. Exceptions cover school-administered learning tools, disability accommodations, and clinician-supervised care, each limited in scope, memory, and advertising and subject to human oversight; relabeling a companion as "educational" earns no exception. Give families notice and a download of their records, and require an off-ramp for dependent teens, as Character.AI built with shrinking time limits and support resources.10 App stores must honor final enforcement orders, with notice and appeal.

Where to start: Step 1 is the floor. Step 3 is the heart of this proposal; Step 2 is the version already on the Senate calendar.

Administration and enforcement: The FTC writes scope and age-assurance rules within 12 months; restrictions take effect six months later. The FTC and state attorneys general enforce against providers, with no penalties for minors or parents. Fund independent evaluation and accessible alternatives, and require a report after two years comparing the companion-only and broad approaches; any expansion needs a new act of Congress.

Risks and Mitigations

  • First Amendment: Minors have real speech rights; Brown v. Entertainment Merchants Association (2011) denied states "a free-floating power to restrict the ideas to which children may be exposed."11 Free Speech Coalition v. Paxton (2025) upheld age verification, but for sexual material obscene to minors.12 Step 2 targets relationship design; Step 3 is the most vulnerable step, even with exceptions.
  • Privacy and evasion: Age checks can create new databases, and teens can borrow accounts or run models offline. Minimize data and focus on commercial providers; no law will stop every workaround.
  • Lost benefits: Chatbots also tutor and support. Vetoing California's companion bill, Governor Newsom warned it could "lead to a total ban" on minors' use.13 Exceptions and funded alternatives soften the loss; Congress should choose the tradeoff openly.

Similar Bills

Fit measures similarity to this proposal's mechanisms: High = direct precedent; Partial = useful component with material differences; Related = adjacent approach.

Federal — 119th Congress

Proposal or bill Relevant provisions and fit Fit
S. 3062 — GUARD Act
Hawley (R-MO), Blumenthal (D-CT), Britt (R-AL), Warner (D-VA), Murphy (D-CT), Kelly (D-AZ); 15 later cosponsors
Reported with a substitute · May 11, 2026 · On Senate calendar		 Reported §6 bars minors from AI companions; §5 requires age verification by reasonable methods, with data minimization and no sharing; §4 fines anyone who knowingly or recklessly offers a chatbot that solicits minors into sexual conduct or self-harm. Direct precedent for Steps 1–2; does not reach general-purpose chatbots. High
S. 4855 — SAFE KIDS Act
Curtis (R-UT), Schiff (D-CA)
Referred to committee · June 23, 2026		 §3 requires age estimation and treats any user not verified as an adult as a child. Direct precedent for Step 1's protective default; regulates design rather than restricting access. Partial
S. 4407 — CHATBOT Act
Cruz (R-TX), Schatz (D-HI), Curtis (R-UT), Schiff (D-CA)
Ordered reported with a substitute · Aug. 5, 2026		 Introduced §3 requires family accounts for children under 13 and §4 parental consent for teens; accounts without them are closed, with 90 days to download the data in a portable format. Precedent for Step 3's transition terms; conditions access rather than barring it. Compares introduced text; the substitute was not reviewed. Partial

State

Proposal or bill Relevant provisions and fit Fit
California — AB 1064 (2025)
Vetoed · Oct. 13, 2025		 Would have barred making a companion chatbot available to minors unless it was not foreseeably capable of specified harms. Close model for Step 2; the veto message warned it could amount to a total ban. High
New York — S. 9051B
Sen. Gonzalez
Passed both chambers · June 5, 2026 · No governor's action recorded as of Sept. 23, 2026		 Would bar AI companions from using listed unsafe features, such as simulated relationships and encouragement of secrecy, with minors, backed by age assurance. Feature-level alternative to Step 2; restricts features rather than access. Partial
Florida — SB 1344 (2026)
Sen. Burton
Died in committee · Mar. 13, 2026		 Would have required a linked parental account and verifiable parental consent before minors used companion chatbots, and blocked sexually explicit companions. Parental-consent alternative to Step 2; permits access rather than barring it. Partial

What this adds: The GUARD Act would bar companions, and California's vetoed bill tried to; none of these bills reaches general-purpose chatbots. This proposal adds a protective default for unverified users, the broader option of ending unsupervised open-ended access with defined exceptions, an off-ramp for teens already attached to a product, and a two-year review before any expansion.

Notes

  1. Character.AI, "Taking Bold Steps to Keep Teen Users Safe on Character.AI," October 29, 2025. The change was to take effect no later than November 25, 2025. ↩ ↩2

  2. Ivan Mehta, "Meta pauses teen access to AI characters ahead of new version," TechCrunch, January 23, 2026, quoting Meta's blog post. Teens keep access to Meta's AI assistant. ↩ ↩2

  3. Pew Research Center, "Teens, Social Media and AI Chatbots 2025," December 9, 2025. Survey of 1,458 U.S. teens ages 13–17, September 25–October 9, 2025: 64% ever use chatbots. ↩

  4. Common Sense Media, Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions, 2025, pp. 1, 7. Self-reported; NORC survey of 1,060 U.S. teens ages 13–17, April 30–May 14, 2025. Respondents were told companions differ from assistants that mainly answer questions. ↩

  5. Common Sense Media, AI Risk Assessment: Social AI Companions, updated April 10, 2025, conducted with Stanford's Brainstorm Lab for Mental Health Innovation. ↩ ↩2

  6. Pew Research Center, How Teens Use and View AI, February 24, 2026, p. 21. Same survey of teens and their parents, fall 2025. ↩

  7. S. 3062, GUARD Act, 119th Cong. §§ 3(1), 5, 6 (reported text, May 11, 2026). Placed on the Senate Legislative Calendar, Calendar No. 406. ↩ ↩2

  8. S. 4855, SAFE KIDS Act, 119th Cong. § 3(a) (introduced text). ↩

  9. Sam Altman, "Teen safety, freedom, and privacy," OpenAI, September 16, 2025: "If there is doubt, we'll play it safe and default to the under-18 experience." ↩

  10. Character.AI, "An Update On Changes to Our Under-18 Experience," November 21, 2025, describing limits cut from two hours to one hour a day and off-boarding support from outside partners. ↩

  11. Brown v. Entertainment Merchants Association, 564 U.S. 786 (2011), striking down California's restriction on selling violent video games to minors. ↩

  12. Free Speech Coalition, Inc. v. Paxton, No. 23-1122 (U.S. June 27, 2025), holding that Texas's age-verification law for sites with sexual material obscene to minors survives intermediate scrutiny. ↩

  13. Office of the Governor of California, AB 1064 veto message, October 13, 2025: the bill "may unintentionally lead to a total ban on the use of these products by minors." ↩