The U.S. Department of Commerce's Center for AI Standards and Innovation (CAISI) recently announced a set of agreements with major artificial intelligence companies — Google DeepMind, Microsoft, and xAI — that would permit government officials to conduct pre-release inspections of advanced AI models. However, the official government webpage describing these partnerships has now vanished from the CAISI website. As of Monday evening, the original URL redirects to the main CAISI landing page, leaving no trace of the original announcement except through archived versions like the Wayback Machine.
Background of AI Vetting Agreements
The agreements, initially signed around May 5, 2026, were part of a broader effort by the National Institute of Standards and Technology (NIST) to evaluate frontier AI capabilities before they reach the public. The program built on earlier partnerships with Anthropic and OpenAI, which were established in 2024 under similar terms. These collaborations were designed to give the government a clearer understanding of leading AI systems' safety, security, and competitiveness in the global AI race.
The announcement described the expanded collaborations as essential for 'information-sharing' and 'ensuring a clear understanding in government of AI capabilities and the state of international AI competition.' The agreements allowed CAISI to conduct targeted research and assessments on unreleased models from Google, Microsoft, and xAI, aiming to advance AI security standards across the industry.
The Missing Announcement
Late last week, reporters and researchers attempting to access the original press release found that the page was returning an error: 'Sorry, we cannot find that page.' Within hours, the URL began redirecting visitors to the main CAISI page, effectively scrubbing the original document from public view. The move was first noted by Reuters on Monday afternoon, though no official explanation has been provided by the White House or the Commerce Department.
Gizmodo reached out for comment but did not receive an immediate reply. The timing of the removal is particularly notable given that the announcement was made only a week prior, and the agreements themselves represent a significant step in government oversight of artificial intelligence. Critics argue that the sudden disappearance undermines transparency and public trust, especially as AI technology becomes more pervasive.
Implications for Transparency and Oversight
The disappearance of such a routine government document has sparked speculation about possible reasons. Some observers suggest that the page may have been taken down inadvertently due to a site migration or technical error. Others point to a potential shift in policy or a desire to control the narrative around government-AI collaborations. Without an official statement, the vacuum of explanation fuels suspicion.
In recent years, the U.S. government has ramped up its involvement in AI safety through executive orders and agency directives. The AI Action Plan, referenced in the original announcement, outlines a strategy for maintaining American leadership in AI while mitigating risks. The vetting deals with Google, Microsoft, and xAI were considered key components of that plan. Their public removal may signal either a revision of those agreements or an effort to limit public scrutiny of the terms.
Furthermore, the lack of a dedicated page makes it harder for researchers, journalists, and the public to track the progress and outcomes of these evaluations. Open government advocates have long stressed the importance of maintaining accessible records of such agreements to ensure accountability. The disappearance of the page also raises questions about the permanence of digital government information and the protocols governing its removal.
Broader Context of AI Policy
The United States is not alone in grappling with how to oversee frontier AI development. The European Union has implemented the AI Act, which mandates certain risk-based requirements for high-risk AI systems. China has also established regulations requiring approval for generative AI models. These international frameworks highlight the growing consensus that proactive government vetting is necessary, yet the methods and transparency vary widely.
The partnerships with companies like Google, Microsoft, and xAI represent a voluntary approach to oversight, as opposed to mandatory regulation. This model has been praised by industry for its flexibility but criticized by civil society for its lack of enforceability. The sudden removal of the announcement could be interpreted as a step back from voluntary transparency, even if the underlying agreements remain in place.
Additionally, the role of NIST and CAISI in AI standards is evolving. NIST has been tasked with developing guidelines for AI risk management, testing environments, and measurement standards. The now-missing page was part of that ongoing effort to formalize government practices. Its loss may hinder collaborative research and public understanding of how government agencies evaluate cutting-edge AI technologies.
Reactions and Next Steps
Industry analysts and watchdogs have called for the Commerce Department to restore the page and provide a clear explanation. Some have filed Freedom of Information Act requests to uncover the reasoning behind the removal and to obtain copies of the agreements themselves. The incident may also spark a broader discussion about the digital transparency of government operations, particularly in areas of national security and emerging technology.
As of now, the Commerce Department has not issued a statement, and the redirect remains in place. The archived version via the Wayback Machine still contains the original text, but relying on unofficial copies is not a sustainable solution for long-term access. The episode serves as a reminder of how quickly official records can vanish in the digital age, and it underscores the need for robust archival practices within federal agencies.
For those tracking AI policy developments, the missing page is a small but telling incident. It may signal operational issues, intentional obfuscation, or simply a bureaucratic slip. Regardless of the cause, the public deserves clarity on such a direct interface between government and major technology firms. The absence of the page does not erase the existence of the agreements, but it does erode the trust that the public can rely on consistent access to information about how their government governs transformative technologies.
Source: Gizmodo News