Americans for Responsible Innovation urged the Trump administration to require security reviews for frontier AI models before public release, warning that advanced systems could accelerate cyberattacks and create national security risks.
The advocacy group said AI developers should have to pass federal reviews to remain eligible for lucrative government contracts, according to Reuters. The proposed checks would focus on whether upcoming models can assist with cyberattacks or weapons development.
The push comes as the White House examines the implications of Anthropic’s Mythos model, which has raised concerns because of its reported ability to make complex cyber operations faster and easier to execute. The debate has shifted AI oversight toward immediate national security threats rather than broader ethics concerns.
The US Center for AI Standards and Innovation already reviews some models through voluntary agreements with OpenAI, Anthropic, Google, Microsoft, and xAI. Americans for Responsible Innovation said CAISI should lead the development of mandatory requirements, while Congress should create a permanent enforcement office inside the Commerce Department.
The proposed requirements would apply to companies spending at least $100 million a year on compute to train frontier models or generating at least $500 million in annual revenue from AI products and services. That threshold would target the largest AI developers while leaving smaller firms outside the initial scope.
The recommendation comes as Washington weighs how to regulate powerful AI systems without slowing US leadership in the sector. Google’s threat intelligence team recently warned that AI powered hacking has moved toward industrial scale, with criminal and state linked actors using commercial models to speed up malicious activity.
