Congressional efforts to create one national rulebook for artificial intelligence are being pulled into a separate fight over children’s online safety, raising the odds that any federal AI bill will become a broader platform-policy bargain rather than a narrow preemption measure.
The immediate issue is preemption: whether Congress should override or limit state AI laws so companies face one federal framework instead of a patchwork of state rules. But recent negotiations have tied that goal to the Kids Online Safety Act, age-verification policy, digital-replica protections, AI chatbot safeguards, and the politics of online speech. That makes the legislation more attractive to some factions in Washington and harder to move as a clean AI governance bill.
The latest round comes after Reps. Jay Obernolte and Lori Trahan released a discussion draft of the Great American AI Act on June 4. Their House proposal is meant to solicit feedback before formal introduction and would build a federal framework for AI governance. Separately, Sen. Marsha Blackburn’s TRUMP AMERICA AI Act framework explicitly links AI policy with children, creators, conservatives, and communities, including pieces of KOSA and the NO FAKES Act.
Why KOSA changes the AI bill
On paper, AI preemption and children’s online safety can overlap. AI chatbots, recommendation systems, engagement features, synthetic media, and age-gated platforms all affect minors. In practice, they are different legislative fights with different coalitions, enforcement models, and tradeoffs.
KOSA-style bills focus on covered platforms and design duties for minors. The Senate version has centered on a duty of care, safeguards for young users, transparency, and age-related study or guidance. Blackburn’s draft AI package goes further by including a dedicated duty-of-care title for AI chatbot developers, a KOSA subtitle, filter-bubble transparency provisions, a GUARD Act section covering AI companions, and a risk-based framework for AI systems, according to the bill text posted by her office.
That bundle creates a practical question for companies: is the prize of national AI preemption worth accepting stricter obligations on youth safety, AI companions, platform design, and digital replicas? For state officials and advocacy groups, the question cuts the other way: would federal law protect young users and consumers strongly enough to justify limiting state experimentation?
What is actually on the table
The House AI draft and Blackburn’s Senate-side framework are not the same bill. The House effort is a bipartisan discussion draft that its sponsors describe as a starting point for feedback on AI governance. Blackburn’s framework is a broader political package that would create a federal AI rulebook and fold in several culture-war and platform-policy priorities.
Blackburn’s summary says the framework would place a duty of care on AI developers, require risk assessments around algorithmic systems and engagement mechanics, add safeguards for users under 17 on covered platforms, create AI-related job-effects reporting, and require data center operators to pay the full cost of needed energy and water infrastructure. The draft bill’s table of contents also includes a prohibition on minor use of AI companions, an advanced AI evaluation program, legal standards for advanced AI products, a National Artificial Intelligence Research Resource, and ratepayer protections for data center infrastructure costs.
That scope explains why the politics are unstable. A single AI-preemption provision could already draw opposition from states with active AI laws and from lawmakers who want local consumer protections. A package that also touches youth safety, speech, Section 230, likeness rights, chatbot design, and energy costs gives almost every stakeholder a reason to ask for changes.
The state-law conflict is not theoretical
The fight matters because state AI rules are already moving into areas such as hiring, housing, lending, health care, education, companion chatbots, deepfakes, automated decision systems, and disclosure duties. Companies building or deploying AI products may prefer one federal standard, but state lawmakers see local rules as a way to respond faster than Congress can.
A broad preemption clause could reshape that map before many state laws are tested. Depending on how Congress writes it, preemption could block only state rules aimed specifically at AI models, reach deployers using AI in high-risk settings, or carve out areas such as privacy, civil rights, consumer protection, labor, and child safety. Those details matter more than the headline phrase “one federal rulebook.”
That is why policy groups are already warning against jamming the issues together. The Information Technology and Innovation Foundation, which supports preempting state AI laws, argued on June 10 that tying AI preemption to children’s safety and digital-replica bills could undermine each debate. The group said Congress should handle AI preemption separately while continuing to work through KOSA, app-store accountability, and NO FAKES concerns.
What platforms and AI companies should watch
For AI companies, the most important provisions to watch are not just model-safety language. Chatbot duty-of-care rules, minor-access restrictions for AI companions, risk assessments, and advanced AI evaluation programs could affect product design, logging, red-teaming, age gates, user controls, and release processes.
For social platforms and app stores, KOSA and age-verification language may matter as much as the AI sections. A bill that imposes design duties for minors could change recommendation systems, default settings, parental controls, notification patterns, advertising practices, transparency reports, and the way platforms treat accounts they believe belong to teenagers.
For developers building consumer AI products, the digital-replica and AI-companion pieces are worth tracking closely. Voice cloning, likeness generation, character chatbots, romantic companion apps, and child-facing AI tools are increasingly being regulated as platform-safety issues, not merely as model-capability issues.
The calendar is the pressure point
The politics are compressed. Congress is already juggling security, budget, defense, immigration, crypto, and election-related priorities, and the summer recess leaves little room for a sprawling tech package. Even supporters of a federal AI framework have to decide whether a broad coalition bill can move faster than a narrower measure.
That makes the next few weeks important. Watch for a formal bill text that reconciles the House AI discussion draft with Blackburn’s framework, signs that House and Senate Republicans agree on which KOSA version to use, and whether Democrats are willing to support any package that limits state AI laws. Also watch for carveouts: the real fight will be in the exceptions that decide which state rules survive.
For now, the story is not that Congress has settled on a national AI law. It is that AI preemption is being traded in a much larger negotiation over how platforms treat children, how AI systems are evaluated, who controls synthetic likenesses, and how much room states will have to regulate the next wave of consumer AI products.
