About a week ago, the U.S. Commerce Department’s Center for AI Standards and Innovation (CAISI) announced a landmark deal with three of the world’s largest AI companies—Google DeepMind, Microsoft, and xAI—allowing the government to inspect unreleased AI models before public deployment. The announcement, dated May 5, 2026, was a significant step in the ongoing effort to ensure that frontier artificial intelligence systems are safe, secure, and aligned with American values and national security interests.
However, as of late Monday evening, that official announcement has vanished from the CAISI website. Attempts to access the original URL first led to an error page stating “Sorry, we cannot find that page,” and later redirected visitors to the main CAISI page on the Commerce Department’s domain. Reuters was the first to report the disappearance, noting the sudden change in government communications regarding AI oversight.
The archived version of the page, preserved by the Wayback Machine, reveals the full text of the announcement. It describes agreements under which “CAISI will conduct pre-deployment evaluations and targeted research to better assess frontier AI capabilities and advance the state of AI security.” The document also explains that “these agreements support information-sharing” and “ensuring a clear understanding in government of AI capabilities and the state of international AI competition.” The partnerships were described as expansions of previously announced collaborations with Anthropic and OpenAI, which had been in place since 2024 and were renegotiated to reflect new directives from the Secretary of Commerce.
The timing of the removal is curious. The original announcement was published just a week ago, on May 5, 2026. By May 11, the URL was already dead. Neither the White House nor the Commerce Department have provided an explanation. Gizmodo reached out for comment on Monday evening but did not immediately receive a response. As of this writing, the URL remains a redirect to the general CAISI homepage, leaving no trace of the original document.
This incident raises significant questions about government transparency in the rapidly evolving field of artificial intelligence. The CAISI, established under the National Institute of Standards and Technology (NIST), was created to spearhead the development of AI standards and to facilitate voluntary agreements with private-sector AI developers. The pre-deployment evaluation program is intended to catch potential harms—such as bias, misinformation, or security vulnerabilities—before they reach the public. The sudden removal of a key announcement undermines that mission and leaves industry watchers, journalists, and the public in the dark.
Behind the scenes, the AI industry has been under increased scrutiny since the release of powerful large language models by OpenAI, Google, and others. The Biden administration had pushed for voluntary commitments in 2023, and subsequent administrations have continued those efforts, with the Trump administration notably signing an executive order in 2025 that prioritized AI innovation while also calling for risk assessments. The CAISI was a product of that executive order, charged with creating a “center of excellence” for AI evaluation.
The agreements with Google DeepMind, Microsoft, and xAI—the latter founded by Elon Musk—were seen as a victory for the administration’s AI Action Plan. They represented the first time the government would gain systematic access to unreleased models from multiple leading firms. However, the disappearance of the announcement suggests either a technical error, a deliberate removal pending revision, or an attempt to downplay the significance of the deals.
Industry observers have noted that the removal could be linked to internal policy disputes—perhaps over the scope of government access or the legal basis for the agreements. Others speculate that the administration might be reconsidering the terms due to pressure from the companies involved. xAI, in particular, has been vocal about what it perceives as overregulation in AI, and Elon Musk has publicly argued that government oversight should focus on “clear and present dangers,” not speculative risks. Meanwhile, Microsoft and Google have each invested billions in AI infrastructure and are wary of regulations that could slow deployment.
Another possible explanation is a simple website reorganization. Government websites frequently undergo updates, and pages sometimes go missing during migration. However, the specific nature of this removal—the URL returning a “not found” error followed by a redirect—suggests a deliberate takedown rather than an accidental deletion. If the page had been moved, a 301 redirect would typically point to the new location. Instead, the redirect leads to the CAISI landing page, implying that the announcement was intentionally unpublished.
The lack of official communication compounds the problem. In an era of heightened concern over AI safety, transparency is paramount. The public and investors both rely on clear, consistent information from regulators. The removal of this page could erode trust in the government’s ability to manage AI risks. It also feeds into a broader narrative of the federal government struggling to keep pace with the private sector in AI oversight.
Historically, NIST has been a trusted source of technical guidance and standards. Its role in AI is to develop benchmarks and conduct evaluations that help companies build safer systems. The CAISI’s pre-deployment evaluations were meant to be a model for how governments and industry can cooperate proactively. The decision to remove the announcement—whether intentional or accidental—risks setting back that collaborative spirit.
For now, the exact status of the agreements with Google DeepMind, Microsoft, and xAI remains unclear. Without the published announcement, there is no official confirmation of the deals. The companies themselves have not commented publicly on the removal. Gizmodo’s requests for comment from the White House and Commerce Department went unanswered as of Monday night. This article will be updated if a response is received.
The episode also highlights the fragility of digital government records. Even when announcements are preserved via third-party services like the Wayback Machine, the official record is what carries legal and policy weight. If the government unilaterally removes information without explanation, it undermines the principle of public accountability. Particularly for an issue as consequential as AI governance, such actions can have ripple effects across industries and international borders.
Other nations are watching closely. The European Union has already passed the AI Act, which includes mandatory transparency requirements for high-risk AI systems. The United States has favored voluntary agreements, but the removal of this announcement could signal a shift in approach—or at least a lack of coordination. Allies and adversaries alike may interpret the disappearance as a sign of instability in U.S. AI policy.
In the meantime, researchers and journalists who rely on official government sources for information about AI regulation are left to speculate. The archived copy of the announcement offers some clues, but it is not legally binding and does not include any contact information or links to supporting documents. The original announcement also included a note about renegotiated partnerships, suggesting that the deals had been updated. Whether those revisions are still in effect is unknown.
This is not the first time a government AI announcement has gone missing. In 2023, a similar page about the White House’s voluntary AI commitments briefly disappeared, only to reappear days later with minor edits. But the scale of this incident is larger: the agreements involve three of the most powerful companies in the world, and the stakes have never been higher. AI systems are now being integrated into critical infrastructure, healthcare, and defense. Pre-deployment testing is a key line of defense against catastrophic failures.
As the story develops, one thing is clear: the removal of the CAISI announcement has created a vacuum of information that will need to be filled. Whether through an official statement, a restored page, or leaked details from inside the Commerce Department, the facts will eventually emerge. Until then, the AI community and the public can only watch the redirect and wonder what happened to the government’s promise of transparency.
Source: Gizmodo News