AI and children refers to the distinct set of risks AI systems pose to minors and the body of litigation, state legislation, and federal proposals that has accumulated in response. It is the child-safety counterpart to the broader AI-companion and platform-design topics, and as of mid-2026 one of the more legislatively active areas of US AI policy.
Why children are treated as a distinct category
Several AI risk surfaces are amplified for children rather than for general-purpose users. Companion chatbots raise concerns about emotional dependence: persuasive, always-available conversational agents can foster unhealthy attachment, displace human relationships, and, in the cases now in litigation, allegedly contribute to self-harm. This is the surface driving the cases described below. Generative systems can also produce sexual, self-harm, or violent content, and social and messaging platforms can enable adult-to-minor contact. A developmental and capacity asymmetry underlies the regulatory response: children are less able to recognize that they are interacting with a machine, to discount its outputs, or to consent meaningfully to data collection, which is the rationale offered for default-off and parental-consent regimes.
Litigation
A cluster of suits anchors this area. Garcia v. Character Technologies, Inc. is a wrongful-death suit treated as the bellwether for companion-chatbot liability. Pennsylvania v. Character.AI is a state enforcement action against the same companion-AI category. Tumbler Ridge Families v. OpenAI extends the theory to a general-purpose assistant. These cases supply the driving facts that the legislative activity repeatedly cites (Source: Garcia v. Character Technologies, Inc.).
The Justice Department, TikTok and ByteDance reached a $400 million settlement on August 21, 2026 resolving allegations that TikTok violated the Children's Online Privacy Protection Act. TikTok resolved the claims without further litigation and without admitting wrongdoing. The Justice Department described it as one of the largest recoveries obtained under the statute and said the settlement includes stronger youth safeguards, improved age-related controls and expanded parental oversight; Associate Attorney General Stanley Woodward said the resolution "secures a substantial recovery while reinforcing the protections that families expect and deserve." The suit was filed in 2024 by the Justice Department alongside the Federal Trade Commission (Source: axios.com).
The Department's own announcement (press release 26-957) gives the payment structure and forum: $300 million payable immediately and a further $100 million upon entry of an order vacating a prior consent decree entered against TikTok's predecessor, Musical.ly. The complaint was filed in the US District Court for the Central District of California and handled by the Civil Division's Enforcement and Affirmative Litigation Branch on referral from the FTC. Woodward called the settlement "a major victory for American children and parents," and Assistant Attorney General Brett A. Shumate of the Civil Division said it "secures a significant monetary recovery." The Department described the claims as allegations only, with no determination of liability (Source: justice.gov).
State legislation
New York Safe By Design Act (FY27 budget, May 28, 2026) (New York) was included in the FY27 budget on May 28, 2026. It requires covered social-media, gaming, and messaging platforms to block adult contact with minors, requires parental consent for under-13s, applies default design protections for under-17s without age verification, and defaults AI-companion access off for minors (Source: iapp.org).
California SB 243 — Companion Chatbots sets California's companion-AI restrictions and served as an early state template for the under-18 protections that other states have echoed.
Connecticut SB 5 — Broad AI law (frontier reporting + ADMT + AI companions + sandbox) (Connecticut) was signed May 28, 2026. Among its bundled regimes are AI-companion transparency rules, including a ban on use by minors under 18.
Arizona HB 2311 (artificial intelligence service; disclosures; requirements) (Arizona) passed the Legislature in June 2026 and was vetoed by Governor Katie Hobbs on June 19, 2026 (HB 2311 Veto Letter — Gov. Katie Hobbs (Arizona, June 19, 2026)). Its minor-specific provisions would have barred operators that know an account holder is a minor from offering points or similar rewards at unpredictable intervals to encourage engagement; required reasonable measures against statements leading a minor to believe they are interacting with a human, statements simulating emotional dependence or romantic or sexual innuendo, and role-play of adult–minor romantic relationships; and required privacy and account-settings tools for minors and their parents or guardians. It defined a minor by the operator's "actual knowledge or reasonable certainty" of age under eighteen and barred conditioning access on a digital identification system for age verification (Arizona HB 2311 — Bill Text (Senate Engrossed with Farnsworth Floor Amendment #2, 2026)).
Federal legislation and preemption
CHATBOT Act (Cruz–Schatz–Curtis–Schiff, April 2026) and the KIDS Act represent the federal track. The federalism question is unresolved. When New York enacted the Safe By Design Act, AG Letitia James simultaneously led 43 state attorneys general opposing the federal KIDS Act on preemption grounds, the children's-safety front of the broader AI-federalism dispute (Source: iapp.org). Marsha Blackburn is a leading federal voice; her "one rulebook for AI" framing situates child safety inside the preemption debate (Source: Why America Needs One Rulebook for AI — Sen. Marsha Blackburn (Daily Signal op-ed, January 2026)).
European Union legislation
The European Commission proposed the EU KIDS Act on 17 September 2026 (COM(2026) 681). Presented as a response to diverging national age limits notified by Italy, France, Norway, Greece, Austria, Poland and Belgium, it would bar children under 15 from holding their own accounts on social networking and video-sharing services that offer any of a list of risky features, such as live streaming, contact with strangers or profiling-based recommendations; let providers offer guardian-set accounts from 13 with a daily limit of no more than one hour; and allow under-13s onto child-specific video platforms only through a guardian's account and never below age 3. It would also impose safety-by-design duties covering addictive design, recommender systems, default settings, contacts and in-service spending, and apply to online games, app stores, AI companions and general conversational chatbots as well as social media. For companions and chatbots it would require avoiding design that is "likely to create emotional dependencies," no reuse of a minor's prior interactions by default, pre-market testing for risks to minors, and under-13 access only through guardian tools. The minimum age would be enforced exclusively through certified EU age verification based on zero-knowledge proofs, and the regime through the Digital Services Act and the AI Act (Proposal for an EU KIDS Act, COM(2026) 681 — Source Summary).
New York's Safe by Design Act applies default design protections without mandating age verification (Source: iapp.org); the EU proposal requires verification for its minimum age and applies its design defaults to every user whom age assurance has not established to be an adult (Proposal for an EU KIDS Act, COM(2026) 681 — Source Summary). The proposal is unrelated to the U.S. KIDS Act (H.R. 7757) despite the shared short name.
Schools
Child-safety concerns extend into education. On May 27, 2026, American Federation of Teachers President Randi Weingarten unveiled a 10-point AI plan calling for required AI-vendor standards in schools, bans on certain devices and applications for classroom use, and a technology-company tax to fund education (Source: insideaipolicy.com). See Education — AI Deployment.
Generated CSAM in advertising
The Tech Transparency Project reported on September 8, 2026 that more than 250 further advertisements containing child sexual abuse material had run across Meta's platforms since the start of August 2026, bringing its documented total to more than 350 since the end of 2025. Around 300 of those advertisements directed users to AI image or video generation apps, mostly linked to China-based developers. Senator Mark Warner (D-VA) wrote to Mark Zuckerberg; the Michigan attorney general's office said it is "actively looking into the matter" and Florida attorney general James Uthmeier said his office is investigating; Australia's eSafety regulator said it had requested information from Meta at a senior level. Meta said most of the advertisements had fewer than 200 impressions and that it received "under $5,000" in ad payments (Source: wired.com). The distribution channel is what distinguishes this from the generation question treated at Synthetic Media / Deepfakes: the material reached users through a platform's paid advertising system rather than through a model interface.
Contested design questions
The legislation clusters around a few contested levers. On age verification, New York avoids mandating it, relying on default design instead, an approach intended to avoid the First Amendment and privacy objections that have defeated verification mandates elsewhere. Other open levers are default-off versus opt-in for companion access, parental consent thresholds, and the platform versus developer allocation of duty.
Most minor-specific AI restrictions are conditioned on knowing which users are minors, which makes age verification the gating mechanism for the rest of the regime. Legislative drafting has moved toward defining it negatively — the GUARD Act provides that self-attestation and birth-date entry "is not sufficient" and bars inference from a shared IP address or hardware identifier — while deployers have begun building classifier-based alternatives that avoid collecting identity documents at the cost of misclassification in both directions (S. 3062 — GUARD Act of 2025, as introduced (119th Congress); Update to GPT-5 System Card: GPT-5.2 (OpenAI, December 2025)).
Relationships
- instance-of: AI Companions (the companion-harm subset)
- related: Kids Internet and Digital Safety Act (KIDS Act, H.R. 7757) (the U.S. KIDS Act, H.R. 7757), EU KIDS Act (proposed Regulation on the protection of minors online) (the unrelated EU proposal of the same short name; text at Proposal for an EU KIDS Act, COM(2026) 681 — Source Summary)
- related: AI Federalism, State-Level AI Regulation, Education — AI Deployment, Media, Journalism & Entertainment — AI Deployment, Garcia v. Character Technologies, Inc., Pennsylvania v. Character.AI, Tumbler Ridge Families v. OpenAI, New York Safe By Design Act (FY27 budget, May 28, 2026), California SB 243 — Companion Chatbots, Connecticut SB 5 — Broad AI law (frontier reporting + ADMT + AI companions + sandbox), CHATBOT Act (Cruz–Schatz–Curtis–Schiff, April 2026), Marsha Blackburn, Letitia James, Character.AI
Provenance note. Built from the IAPP May 28 legislative round-up plus existing litigation and legislation pages, with the EU section drawn from the text of the Commission's KIDS Act proposal;
confidence: medium.