The Trump administration is currently paralyzed by internal conflict regarding federal AI regulation, with the Commerce Department, intelligence agencies, and pro-industry factions struggling for control. This political deadlock reached a critical point after Anthropic's Mythos model demonstrated the ability to autonomously identify more than 10,000 zero-day vulnerabilities across major operating systems and web browsers. National security officials now argue that AI systems with such offensive cybersecurity capabilities require rigorous evaluation by intelligence agencies, while the Commerce Department maintains that civilian testing is necessary to ensure continued industry cooperation.

The regulatory gridlock became apparent in May 2026. On May 5, the Commerce Department’s National Institute of Standards and Technology (NIST) announced pre-deployment testing partnerships with Google DeepMind, Microsoft, and Elon Musk’s xAI. However, the announcement was removed from the agency's website days later without explanation. Tension further escalated on May 21 when the Trump administration abruptly scrapped a planned executive order intended to formalize government safety vetting for frontier AI models. Trump reportedly canceled the signing ceremony due to concerns that such regulations could hinder American technological leadership.

The absence of a federal framework has created a regulatory vacuum, leaving the United States without formal oversight mechanisms sixteen months after the administration repealed President Biden's original AI executive order. While the European Union’s AI Act is set for full enforcement in August 2026, the US government currently lacks statutory authority to vet powerful models before they are released. This inconsistency has forced some US states to draft independent legislation, creating a fragmented regulatory environment for companies.

Within this impasse, the Office of the National Cyber Director has proposed establishing a centralized AI evaluation center under the Office of the Director of National Intelligence. This move faces pushback from Commerce officials, who fear that integrating AI testing into the intelligence community will be perceived internationally as weaponizing safety protocols. As the administration debates its path forward—weighing a weakened executive order against a limited voluntary commitment model—private sector initiatives like Anthropic’s Project Glasswing are currently filling the void by providing vetted organizations restricted access to Mythos for security testing.

The Trump administration is currently paralyzed by internal conflict regarding federal AI regulation, with the Commerce Department, intelligence agencies, and pro-industry factions struggling for control. This political deadlock reached a critical point after Anthropic's Mythos model demonstrated the ability to autonomously identify more than 10,000 zero-day vulnerabilities across major operating systems and web browsers. National security officials now argue that AI systems with such offensive cybersecurity capabilities require rigorous evaluation by intelligence agencies, while the Commerce Department maintains that civilian testing is necessary to ensure continued industry cooperation.

The regulatory gridlock became apparent in May 2026. On May 5, the Commerce Department’s National Institute of Standards and Technology (NIST) announced pre-deployment testing partnerships with Google DeepMind, Microsoft, and Elon Musk’s xAI. However, the announcement was removed from the agency's website days later without explanation. Tension further escalated on May 21 when the Trump administration abruptly scrapped a planned executive order intended to formalize government safety vetting for frontier AI models. Trump reportedly canceled the signing ceremony due to concerns that such regulations could hinder American technological leadership.

The absence of a federal framework has created a regulatory vacuum, leaving the United States without formal oversight mechanisms sixteen months after the administration repealed President Biden's original AI executive order. While the European Union’s AI Act is set for full enforcement in August 2026, the US government currently lacks statutory authority to vet powerful models before they are released. This inconsistency has forced some US states to draft independent legislation, creating a fragmented regulatory environment for companies.

Within this impasse, the Office of the National Cyber Director has proposed establishing a centralized AI evaluation center under the Office of the Director of National Intelligence. This move faces pushback from Commerce officials, who fear that integrating AI testing into the intelligence community will be perceived internationally as weaponizing safety protocols. As the administration debates its path forward—weighing a weakened executive order against a limited voluntary commitment model—private sector initiatives like Anthropic’s Project Glasswing are currently filling the void by providing vetted organizations restricted access to Mythos for security testing.