Senator Budd Urges AI Safety Center to Resume Public Research

Senator Ted Budd is pressing the White House to allow a federal AI research center to resume publishing its evaluations of advanced artificial intelligence systems, arguing that transparency serves America's competitive interests.

In a June 30 letter to National Cyber Director Sean Cairncross and Office of Science and Technology Policy Director Michael Kratsios, the North Carolina Republican called for the Center for AI Standards and Innovation (CAISI) to continue making its frontier model assessments public. The request follows reports that CAISI has been directed to stop publishing its findings, according to details first reported by Senator Budd's office.

Recent policy shift

The change comes after President Trump's June 2 executive order on AI innovation and security, which tasked the Treasury Secretary, NSA Director, and CISA Director with developing a classified benchmarking process for advanced AI models. That order empowers the NSA director to determine when assessment results should be shared with developers and researchers.

CAISI, housed within the National Institute of Standards and Technology, has established partnerships with leading AI laboratories to evaluate capabilities and identify vulnerabilities. The center has published research highlighting security flaws in Chinese AI models, including DeepSeek's open-source offerings.

Why it matters

The tension between classification and transparency reflects a fundamental challenge in AI governance: how to protect national security while maintaining the open research culture that has driven American AI leadership. Public benchmarks allow the broader research community to build on government findings, help companies make informed decisions about which models to deploy, and expose risks in foreign AI systems that might otherwise go unnoticed by private-sector users.

Balancing security and innovation

In his letter, Budd acknowledged the legitimate security concerns driving the classification push, including risks of technology theft by adversaries and potential attacks through prompt injection or data poisoning. He emphasized that NSA's cybersecurity expertise is critical for protecting infrastructure.

However, the senator argued that CAISI's collaborative research model expands the capacity of non-governmental organizations and companies to monitor AI developments. "More high-quality research conducted by CAISI's informed experts deepens America's AI competitive ecosystem for model and benchmark developers," Budd wrote.

The senator specifically noted CAISI's role in publicly documenting the capabilities and vulnerabilities of Chinese AI models relative to U.S. systems, information that helps researchers and businesses avoid unsafe platforms.

Budd requested that CAISI be allowed to continue its mission alongside other government AI initiatives and that its research be published "to the extent possible." He framed the request as essential for maintaining American leadership as AI adoption accelerates across economic sectors.

Details of the letter and the reported directive to halt publications were first disclosed by Senator Budd's office.