Congress Unveils Plan for Federal AI Watchdog with New Safety Rules

A bipartisan group of House lawmakers has released a draft bill that would create the first federal agency dedicated to overseeing artificial intelligence, marking a concrete step toward regulating a technology that has outpaced existing laws.

The proposal, made public on June 4, 2026, would require companies building advanced AI systems to disclose how their models are trained, what data was used, and what safety checks were performed before release. Developers whose systems exceed certain computing power thresholds would face mandatory transparency rules.

At the heart of the bill is a new government body focused solely on AI governance. This agency would set enforceable standards for high-risk uses such as medical diagnosis, loan approvals, and infrastructure management. Unlike previous voluntary guidelines, this framework includes penalties for violations.

The legislation draws from the European Union's AI Act but adapts it for American legal and economic priorities. It explicitly aims to keep AI development on U.S. soil by avoiding rules that could drive companies overseas. Intellectual property protections are strengthened, with training on public data generally treated as fair use while creators gain new tools to opt out of having their work used in training sets.

Companies would need to report significant AI failures or security breaches within 72 hours, similar to existing cybersecurity notification laws. The bill also funds retraining programs for workers whose jobs face disruption from automation.

Privacy protections include bans on using sensitive personal data in commercial AI training without explicit consent. National security provisions call for export controls on advanced AI chips and increased investment in defending American systems from theft or attack.

The draft gives existing agencies new responsibilities where appropriate. The Federal Trade Commission would police deceptive AI-generated content like deepfakes, while the FDA would oversee medical AI applications.

Small businesses and academic researchers receive exemptions for smaller projects, plus grant programs to help with compliance costs. The bill phases in requirements over one to five years to give industry time to adapt.

Industry groups acknowledged the need for regulation while raising concerns about specific provisions. Civil society organizations welcomed the transparency measures but pushed for stronger protections against bias and privacy violations.

The discussion draft is just the opening move in what promises to be a lengthy legislative process, with additional hearings and negotiations expected before any final bill reaches a vote.