The United States Department of Commerce’s Center for AI Standards and Innovation (CAISI) recently announced a series of agreements with major artificial intelligence companies, including Google DeepMind, xAI, and Microsoft. These deals were intended to allow government officials to preview and evaluate unreleased AI models before they reached the general public. The announcement, dated May 5, 2026, was seen as a significant step toward ensuring that frontier AI capabilities are vetted for safety and security risks. However, within days, the official webpage containing that announcement disappeared from the CAISI website. As of this writing, the original URL redirects visitors to the main CAISI landing page, leaving many to wonder what happened and why.

The missing page was first noticed by Reuters on Monday afternoon. When attempting to access the original link, users encountered an error message stating, “Sorry, we cannot find that page.” Later, the URL was redirected to the CAISI homepage on the Commerce Department site. The disappearance has raised eyebrows among transparency advocates and industry observers, especially given the high-profile nature of the companies involved and the sensitive topic of government oversight of artificial intelligence.

What the Archived Announcement Revealed

Thanks to the Wayback Machine, the original announcement is still accessible. It explained that CAISI, operating under the National Institute of Standards and Technology (NIST), had negotiated expanded industry collaborations with Google DeepMind, Microsoft, and xAI. These agreements built on previously announced partnerships that had been renegotiated to align with directives from the Secretary of Commerce and America’s AI Action Plan. The archived text stated: “These agreements support information-sharing, ensuring a clear understanding in government of AI capabilities and the state of international AI competition.”

The announcement also emphasized that CAISI would conduct pre-deployment evaluations and targeted research to better assess frontier AI capabilities and advance the state of AI security. Such evaluations are crucial in a landscape where AI models are becoming increasingly powerful and opaque. By gaining early access to unreleased models, government researchers can identify potential vulnerabilities, biases, or misuse risks before the technology is widely deployed.

Previous Agreements and the Broader Context

The deals with Google, xAI, and Microsoft follow similar arrangements made with Anthropic and OpenAI back in 2024. Those earlier agreements were hailed as a model for public-private partnership in AI safety. The extension to three more major players indicates that the government is trying to broaden its oversight net, especially as competition in AI development intensifies between the United States and countries like China. The inclusion of xAI, Elon Musk’s AI venture, is particularly noteworthy given Musk’s outspoken views on AI risk and his sometimes adversarial relationship with regulators.

Pre-deployment evaluations are a cornerstone of the Biden administration’s AI policy framework, which emphasizes proactive risk management rather than reactive regulation. However, the sudden removal of the announcement page could signal a shift in strategy or internal disagreements about the transparency of such agreements. Some experts suggest that the government might have removed the page due to concerns over revealing sensitive operational details, such as the specific scope of evaluations or the timeline of model releases. Others speculate it could be a simple technical glitch—though the redirection suggests intentional action.

Why the Disappearance Matters

The missing page is more than a bureaucratic hiccup. It undermines the trust that the public and industry place in government oversight. If the very announcement of a transparency initiative disappears without explanation, it sends a mixed message about the government’s commitment to open governance. The Commerce Department and the White House have not yet responded to requests for comment. As of Monday evening, Gizmodo’s inquiries went unanswered, and no official statement has been issued.

The incident also raises questions about the durability of AI safety commitments. In recent years, multiple tech companies have signed voluntary pledges to allow government audits, but enforcement mechanisms remain weak. The removal of a simple webpage could be a harbinger of more significant policy rollbacks. Moreover, the AI community closely watches these agreements as benchmarks for global standards. The United States has been pushing for international cooperation on AI safety, but if domestic announcements vanish without trace, credibility suffers.

Technical and Political Possibilities

From a technical standpoint, webpage disappearances can occur due to content management errors, accidental deletions, or routine site maintenance. However, the fact that the URL now redirects—rather than simply returning a 404 error—suggests deliberate reconfiguration. It is possible that CAISI is updating the announcement to include additional companies or revised terms, but no such update has appeared. Alternatively, the page might have been taken down due to legal concerns, perhaps because the agreements contained confidential clauses that were inadvertently disclosed.

Politically, the timing is sensitive. The AI Action Plan referenced in the announcement is a key component of the current administration’s technology agenda. Any perceived inconsistency could be seized upon by critics who argue that the government is either too cozy with big tech or too secretive about its regulatory activities. The disappearance of a page that explicitly touted information-sharing and transparency is particularly ironic and could fuel accusations of hypocrisy.

The Significance of Pre-Deployment Evaluations

Pre-deployment evaluations are a relatively new tool in the regulatory arsenal. They allow researchers to test AI systems in controlled environments before they are released to millions of users. This can help prevent catastrophic failures, such as biased algorithms, privacy breaches, or autonomy-related accidents. In the case of large language models and generative AI, evaluations often focus on “alignment” with human values, robustness to adversarial prompts, and potential for misuse in creating disinformation or cyberattacks.

Companies like Microsoft and Google have their own internal safety teams, but independent government scrutiny adds an extra layer of accountability. The agreements signed with CAISI reportedly include provisions for regular check-ins and data sharing. Without access to the original page, however, the exact terms remain unclear. The archived text provides only a high-level summary, leaving many details to speculation.

International Competition and AI Safety

The announcement also referenced “the state of international AI competition,” highlighting that these evaluations are not just about safety but also about maintaining technological leadership. The United States and China are locked in a race to develop the most advanced AI systems, and governments on both sides are seeking to balance innovation with control. By evaluating models early, American officials can gain intelligence on emerging capabilities and potentially steer development toward beneficial uses.

However, critics argue that such collaboration blurs the line between oversight and industrial policy. If the government becomes too involved in pre-release testing, it could inadvertently favor certain companies or steer the direction of research. The fact that the announcement page is now hidden only amplifies those concerns. Transparency is essential for maintaining public trust, and the absence of an official record suggests that even the narrative around these deals is being carefully managed.

The missing page is a small but telling example of how quickly information can be erased in the digital age. In an era where every byte is supposedly preserved, the intentional removal of a government announcement about AI oversight sends a troubling signal. It may be a simple bureaucratic decision, or it may indicate deeper anxieties about how these agreements are perceived. Either way, the incident calls for a prompt explanation from the authorities. As the investigation continues, the public deserves to know why a page that once stood as a symbol of accountability has vanished without a trace.

Source: Gizmodo News