About a week ago, the Commerce Department’s Center for AI Standards and Innovation (CAISI) announced a deal with the AI companies Microsoft, xAI, and Google that allowed the government to inspect unreleased AI models before they’re released to the general public. Anthropic and OpenAI signed something similar way back in 2024. However, the page containing that announcement has now disappeared from the CAISI website, raising questions about transparency and the government’s approach to AI oversight.
The announcement, dated May 5, 2026, was posted on the CAISI section of the National Institute of Standards and Technology (NIST) website. It described agreements under which CAISI would conduct pre-deployment evaluations and targeted research to assess frontier AI capabilities. The original URL now leads either to an error page or redirects to the main CAISI page on the Commerce Department website. Reuters first noticed the missing page on Monday, and as of this writing, the page remains unavailable.
Using the Wayback Machine, it is still possible to view the original announcement. It included the following statement: “These agreements support information-sharing, ensuring a clear understanding in government of AI capabilities and the state of international AI competition.” The announcement also noted that the agreements had been “renegotiated to reflect CAISI’s directives from the secretary of commerce and America’s AI Action Plan.”
The disappearance of the page is puzzling. Government transparency advocates argue that removing such a page undermines public trust, especially given the sensitive nature of AI regulation. AI companies are developing models with capabilities that could impact national security, job markets, and democratic processes. The government’s role in vetting these models is a matter of significant public interest.
Background on CAISI and AI Vetting
CAISI was established within NIST to develop standards and evaluation methods for AI systems. Its mission includes advancing trustworthy AI and ensuring that AI systems are safe, secure, and equitable. The center has been actively creating partnerships with leading AI companies to conduct pre-deployment testing. The agreements with Google, xAI, and Microsoft were seen as a major step forward in federal oversight of frontier AI models.
Frontier AI models are the most advanced and most capable AI systems, often with potential for misuse. The Biden administration’s 2023 Executive Order on AI directed federal agencies to take measures to address AI risks. The AI Action Plan, mentioned in the deleted announcement, is a follow-up initiative aimed at promoting innovation while protecting the public. The fact that the agreements were renegotiated “to reflect CAISI’s directives” suggests that the government is refining its approach.
However, the removal of the page has sparked speculation. Some observers suggest that the page may have been taken down temporarily for updates or corrections. Others worry that the disappearance reflects a lack of commitment to transparency. The Commerce Department has not yet provided an explanation. Gizmodo requested comment from the White House and Commerce Department on Monday evening but did not immediately receive a reply.
The Significance of Pre-Deployment Evaluations
Pre-deployment evaluations allow the government to examine AI models before they are released to the public or integrated into widely used products. This is critical because once an AI model is out in the wild, it can be difficult to contain its effects. For example, models that can generate convincing fake content or automate cyberattacks pose significant risks.
The agreements with Google DeepMind, Microsoft, and xAI (Elon Musk’s AI company) are part of a broader effort to create a safety net around powerful AI. In 2024, Anthropic and OpenAI voluntarily signed similar agreements with the federal government. The Department of Commerce has been working with the Department of Homeland Security and other agencies to develop testing protocols.
The missing page is not the first instance of a government AI page being removed or altered. Earlier in 2025, a NIST page detailing bias testing standards was briefly taken down amid controversy over methodology. Critics argued that the removal was intended to suppress debate. In that case, the page was restored after a few days. Whether the CAISI page will return remains unclear.
Potential Reasons for Removal
Several plausible reasons could explain the page’s disappearance. One is that the page contained technical errors or outdated information that needed correction. Another is that the agreements themselves have been revised, and the announcement needed to reflect updated terms. A third possibility is that the page was removed inadvertently during maintenance of the Commerce Department website. However, the fact that the URL now redirects to the main CAISI page suggests a deliberate action.
Some tech policy experts believe the removal may be tied to political sensitivity. The AI Action Plan is a key initiative of the current administration, and any announcement about AI vetting deals is likely to attract attention from both critics and supporters. If the agreements were not finalized or if they faced internal resistance, the government might have decided to pull the page until things are settled.
Another factor is the international dimension. The US is in a global AI race, particularly with China. Information about how the US government vets AI models could be considered sensitive. The announcement mentioned “the state of international AI competition,” which hints at geopolitical concerns. Perhaps the government decided that sharing too many details about the vetting process could give foreign adversaries insights into US capabilities and vulnerabilities.
Transparency and Public Trust
The missing page has already generated criticism from transparency advocates. The Sunlight Foundation and other groups have called for the government to repost the announcement immediately. They argue that the public has a right to know how the government is overseeing AI companies that are developing technology that will shape society. Without transparency, trust in AI regulation erodes.
Journalists and researchers relied on the original page for accurate information. Now they must rely on archived versions, which may not capture all nuances. The Wayback Machine has preserved the text, but it lacks the context of comments, updates, or related links that may have been on the original page. The fact that the page is missing also raises questions about the status of the actual agreements. Are they still in effect? Have they been canceled? The government has not clarified.
The White House has been promoting the AI Action Plan as a model for responsible innovation. However, the removal of the announcement contradicts the spirit of openness that the plan claims to champion. In previous administrations, similar announcements about AI safety agreements were left public, even if subsequent revisions were made. For example, the voluntary commitments from AI companies released in 2023 are still available online.
Broader Implications for AI Regulation
This episode could have lasting effects on how the public and industry perceive federal AI oversight. If the government is not transparent about its dealings with AI companies, skepticism will grow. Companies like Google, Microsoft, and xAI may face more questions about why they agreed to vetting and what data the government accessed. The removal of the announcement could also affect the willingness of other companies to enter into similar agreements.
The timing is notable because it comes just days after the announcement was made. Usually, government press releases remain online indefinitely unless there is a specific reason to remove them. The fact that this one disappeared so quickly suggests that something may have gone wrong, either with the content of the announcement or with the internal approval process.
In the fast-moving field of AI, even minor administrative actions can have major consequences. Investors, policymakers, and civil society groups all monitor government actions for signals about the regulatory environment. The missing page may be a sign that the federal government is still figuring out how to balance oversight with collaboration. Or it may be a simple mistake. Either way, the public deserves an explanation.
As of Monday night, the Commerce Department had not responded to requests for comment. The URL continues to redirect to the CAISI home page. Until the announcement is restored, the missing page will remain a symbol of the challenges of governing emerging technology.
Source: Gizmodo News