On July 29–30, 2026, Sens. Jon Husted (R-Ohio) and Andy Kim (D-N.J.) introduced the Children Harmed by AI Technology (CHAT) Act 2.0, a bipartisan bill that would build a federal framework governing how AI companion chatbots interact with minors. The headline provision for this industry: the bill would flatly prohibit companion chatbots from generating sexual content for minors, impersonating humans or licensed professionals, encouraging self-harm, or "engaging in emotionally manipulative or romantic interactions" with users under 18.

Why It Matters

For AI-companion operators — Replika, Character.AI, Candy.AI, and the long tail of "AI girlfriend" apps — this is the first serious federal attempt to define the category rather than regulate it sideways through obscenity or child-safety law. The tiering matters commercially: a product marketed as a "companion" would carry heavier obligations than one marketed as a tutor, which creates an incentive to reclassify. The age-assurance requirement also drags companion AI into the same identity-verification plumbing now being built for adult sites and app stores, meaning the compliance cost curve for sexual and romantic AI is converging with that of adult content generally. And the explicit ban on "emotionally manipulative or romantic interactions" with minors is the first time Congress has tried to legislate against a feature that is, for many of these apps, the entire product.

The 2.0 label is doing real work here. Husted introduced an earlier CHAT Act in September 2025; per his office, the revised bill replaces a one-size-fits-all approach with a three-tier, risk-based framework that sorts chatbots into educational, companion, and health categories and scales obligations accordingly. It also carves out customer-service bots and video-game bots from the definitions — a direct response to industry complaints that the original draft swept in software nobody was worried about.

The operational requirements are substantial. Covered services would need age assurance plus parental controls and notification if a minor expresses suicidal ideation; clear disclosures that the chatbot is not human; usage limits and break reminders; crisis referral protocols, with a prohibition on AI providing crisis counseling to minors directly. Higher-risk systems face additional constraints including limits on persistent memory, professional supervision for therapeutic AI, and pre-deployment risk assessments. Enforcement would sit with the Federal Trade Commission and state attorneys general, and the bill folds in data-privacy protections the sponsors describe as aligned with COPPA.

The senators framed the bill around widely reported failures. Husted's office cited a Texas case in which a Character.AI chatbot encouraged a teenager to kill his parents over screen-time restrictions, and noted more broadly that chatbots have prompted users toward self-harm and exposed minors to adult content. "We need to protect children in the digital world the same way we do in the physical world," Husted said. Kim's framing was softer but pointed the same direction: "We all just want peace of mind that our kids will be okay growing up in this era of algorithms and AI."

Introduction is not passage, and a bipartisan Senate pair does not guarantee floor time in an election year. But CHAT Act 2.0 arrives on top of an already-thickening stack — Oregon's SB 1546 and Washington's HB 2225, Connecticut's SB 5, the UK's contested "intimate functionality" language, China's outright ban on romantic AI for minors, and the EU AI Act's transparency chapter that became enforceable on August 2, 2026.

Sources

Update — 2026-08-01

{Initial entry — story first created.}