White House Cyber Office Clashes With Commerce's CAISI Over AI Model Evaluation Authority

A turf battle is brewing inside the federal government over who gets to set the rules for judging whether artificial intelligence models are safe and secure. Sources say the White House's Office of the National Cyber Director and the Commerce Department's AI Safety Institute are locked in a standoff over which agency should lead AI model evaluations—a fight that underscores how Washington is still struggling to draw clear lines of authority over emerging AI technology.

The core of the dispute centers on competing visions for AI oversight. Intelligence officials want a stronger hand in shaping AI policy that currently falls under Commerce's purview, according to people familiar with the matter. The ONCD, established just a few years ago to coordinate federal cybersecurity strategy, has been pushing to expand its influence over AI-related risk assessment. Meanwhile, Commerce's CAISI was built specifically to evaluate AI systems for safety and national security risks, giving it a technical edge but perhaps less political leverage.

The clash matters because unclear authority can slow down or complicate federal responses to AI risks. Companies, researchers, and foreign partners looking to the U.S. government for guidance on AI standards could face conflicting signals if two agencies are pulling in different directions. The fight also reflects a broader pattern across government: as AI capabilities advance faster than institutional structures can adapt, competing bureaucratic interests are rushing to fill the vacuum. Resolving who leads evaluations will shape how rigorously AI models are vetted before and after deployment.