A watchdog organization is pushing the Trump administration to implement mandatory security evaluations for advanced artificial intelligence systems before companies can secure federal contracts.
Americans for Responsible Innovation submitted recommendations to government officials on Monday, calling for comprehensive screening processes that would examine AI technology for potential cybersecurity threats and weapons development risks before public deployment.
The proposal comes as the White House confronts concerns about Anthropic’s Mythos system, which experts worry could streamline sophisticated cyber operations and create national security vulnerabilities.
Under the advocacy group’s plan, AI developers would need to successfully complete security assessments to qualify for government contracting opportunities. The organization emphasized that these evaluations should focus on preventing the creation of tools that could facilitate cyberattacks or military applications.
Currently, the U.S. Center for AI Standards and Innovation conducts voluntary assessments of certain AI systems through partnerships with major technology companies including OpenAI, Anthropic, Google, Microsoft and xAI.
The advocacy group recommends expanding CAISI’s role to include developing binding standards, while suggesting Congress establish a dedicated enforcement division within the Department of Commerce to oversee compliance.
The proposed regulations would target companies investing at least $100 million annually in computational resources for training advanced AI models, or those generating $500 million or more yearly from artificial intelligence products and services.
This framework mirrors safety reporting standards that California implemented in the previous year, establishing similar financial thresholds for regulatory oversight.
