#What Changes Have Occurred with AI Model Oversight?
The Center for AI Standards and Innovation has recently ceased the public reporting of its findings regarding AI model assessments, following a new executive order issued on June 2, 2026. This directive transfers oversight authority from CAISI to a classified system managed by national security agencies.
This shift was precipitated by the introduction of Anthropic's unreleased Claude Mythos model, which demonstrated an extraordinary ability to autonomously detect thousands of previously overlooked high-severity zero-day vulnerabilities within major operating systems. The implications of this move are significant for the technology landscape and investors in AI firms.
#Why Does This Change Matter?
By early June 2026, CAISI had completed evaluations for over 40 AI models, establishing a valuable public repository of knowledge that benefited researchers, policymakers, and the tech sector. The new executive order requires a comprehensive 30-day governmental review of AI models prior to their public release. This evaluative authority will now rest with three key agencies: the Treasury Department, the National Security Agency, and the Cybersecurity and Infrastructure Security Agency.
The sophistication of Anthropic’s Mythos prompted this decision, indicating that existing oversight mechanisms not only failed to prevent vulnerabilities but were seen as inadequate to address the risks posed by rapidly evolving AI capabilities.
#How Will This Affect AI Companies and Investors?
This revised protocol introduces a new source of friction for major AI entities like Anthropic, OpenAI, and Google DeepMind, all of which must incorporate government compliance into their release schedules. The previous open evaluations provided a transparent benchmarking framework, which is now significantly altered. Companies will still submit their models for review, yet evaluation results will largely remain confidential, stripping away independent verification.
Consequently, US-based AI companies are now subject to a verification process that their international counterparts, particularly in China and Europe, are not obliged to follow. This disparity could create competitive disadvantages in the global market. While some researchers affiliated with CAISI may adapt to this new paradigm of national security, others are likely to transition to academia or private sectors that still permit open scientific publication.