The United States is racing to put guardrails on a new class of cyber-capable artificial intelligence models before China closes the gap, with the Trump administration and a bipartisan duo in the House moving in the same week to shape how frontier systems reach the public.

Recent estimates cited by Politico put Washington's window at six to 12 months before Beijing fields a comparable model or develops one of its own.

President Donald Trump on June 2 signed an executive order titled "Promoting Advanced Artificial Intelligence Innovation and Security," directing federal agencies to build a voluntary framework under which developers of "covered frontier models" may give the government up to 30 days of prerelease access to inspect cybersecurity risks.

The order bars any mandatory licensing or preclearance regime and tasks the National Security Agency, the Cybersecurity and Infrastructure Security Agency, and the Treasury Department with developing a classified benchmark to identify which models qualify.

The order arrives after Anthropic's Claude Mythos and OpenAI's GPT-5.5-Cyber, both unveiled in April, demonstrated an ability to autonomously find software vulnerabilities at machine speed.

Anthropic has said it is sharing Mythos with roughly 150 organizations across 15 countries under its Project Glasswing program, conditioned on undisclosed security requirements.

Two days later, Reps. Jay Obernolte, R-Calif., and Lori Trahan, D-Mass., released a 269-page discussion draft titled the Great American Artificial Intelligence Act of 2026.

The bill would impose binding transparency, audit, and whistleblower requirements on "large frontier developers," defined as companies with more than $500 million in gross revenue in the prior year, while barring states for three years from passing laws specifically regulating how AI models are developed.

State laws governing the use or deployment of AI systems are not preempted, though critics argue the development carveout is where state oversight matters most.

The preemption provision drew immediate pushback from labor unions, consumer advocates, and the House Democratic Commission on AI and the Innovation Economy, complicating the draft's path even as supporters argue the three-year window is narrower than a 10-year moratorium Congress rejected last year.

The administration has not said whether it will back the measure.

Anthropic, in a May 14 policy paper titled "2028: Two Scenarios for Global AI Leadership," argued that the United States and its allies must "stay ahead of authoritarian governments like the Chinese Communist Party," warning that AI could soon be used "to repress citizens at unprecedented scale."

The company called for tighter export controls on advanced chips and steps to curb distillation attacks, in which smaller models are trained on the outputs of frontier systems.

Beijing has signaled its ambitions by embedding AI in its next five-year plan and preparing to host a world AI summit in Shanghai in July.

The Chinese Embassy in Washington told Politico that AI "should not be owned by major countries, still less dominated by contests and rivalry."

Whether the U.S. lead holds, several cybersecurity executives told Politico, will turn on how quickly defenders can use the same tools to patch critical systems before attackers reach for them.