In a move that has raised eyebrows among technology policy experts, the U.S. Department of Commerce has apparently removed a web page that announced unprecedented agreements allowing government officials to inspect unreleased artificial intelligence models from some of the world’s leading AI companies. The page, which belonged to the Center for AI Standards and Innovation (CAISI) within the National Institute of Standards and Technology (NIST), had celebrated new partnerships with Google DeepMind, Microsoft, and xAI—the AI venture founded by Elon Musk.

The announcement was dated May 5, 2026, but by the following week, the original URL had begun redirecting to the main CAISI landing page, displaying a “Sorry, we cannot find that page” error before the redirect was implemented. The disappearance was first noted by Reuters and later confirmed by other outlets. As of this writing, the page remains inaccessible via its original link.

This development occurs against a backdrop of heightened scrutiny around AI governance. The agreements, which built on earlier pacts with Anthropic and OpenAI signed in 2024, were intended to allow CAISI to conduct pre-deployment evaluations and targeted research on frontier AI models. The goal, according to the archived text retrieved via the Wayback Machine, was to “support information-sharing” and “ensure a clear understanding in government of AI capabilities and the state of international AI competition.”

The Agreements and Their Significance

Frontier AI models—large language models and multimodal systems trained on massive datasets—have become central to both economic competitiveness and national security. The agreements with Google DeepMind (the Alphabet subsidiary responsible for advanced AI like Gemini and AlphaFold), Microsoft (which has integrated OpenAI’s technology into its Azure cloud platform and consumer products), and xAI (which developed the Grok chatbot and has ambitious plans for data centers) would have given the U.S. government a unique window into the inner workings of these models before public release.

Such pre-deployment vetting is seen by many policymakers as essential to mitigate risks such as bias, misinformation, cybersecurity vulnerabilities, and even existential threats. The Biden administration had previously issued an executive order on AI safety, and the Trump administration—as of 2026, Donald Trump is back in office—has continued with its own “America’s AI Action Plan,” which explicitly directs CAISI to renegotiate and expand these industry collaborations.

The May 5 announcement stated that CAISI would conduct “pre-deployment evaluations” and engage in “targeted research to better assess frontier AI capabilities and advance the state of AI security.” It stressed that the agreements were renegotiated to align with the directives from the Secretary of Commerce and the new AI Action Plan. The language suggested a strengthening of government oversight, a message that now seems at odds with the quiet removal of the page.

Possible Reasons for the Removal

Speculation has emerged about why the page was taken down. Some observers suggest it could be a routine website restructuring—perhaps the content was merged into a broader policy page. Others point to potential political sensitivity. The 2026 political landscape is fraught with debates over AI regulation, with some Republicans favoring lighter-touch oversight to foster innovation, while Democrats call for more stringent guardrails. The removal might indicate internal disagreements within the administration about how much to publicize these agreements.

Another possibility is that the page contained technical details or language that the companies objected to, leading to a quiet correction. Google, Microsoft, and xAI have all invested heavily in AI safety teams, but they also compete fiercely. Publicly disclosing the exact scope of government access could impact competitive intelligence or market perceptions.

There is also the question of timing. The page went missing just days after the announcement, during a week when other AI-related news dominated headlines—including a planned strike at Samsung’s semiconductor plants that could disrupt AI chip supply chains, and OpenAI’s unveiling of “Daybreak,” its answer to Anthropic’s Project Glasswing. The disappearance may have been overlooked by many, but for those tracking government transparency, it is a concerning signal.

Historical Context of Government AI Oversight

The U.S. government has a mixed record on overseeing emerging technologies. During the early days of the internet, minimal regulation allowed explosive growth but also enabled problems like misinformation and privacy violations. With AI, policymakers are attempting a more proactive approach, but the rapid pace of development often outstrips bureaucratic processes.

In 2023, the Biden administration issued an executive order requiring developers of powerful AI models to share safety test results with the government. The order also established the U.S. AI Safety Institute at NIST, which later evolved into CAISI. These moves were seen as a global benchmark, influencing the European Union’s AI Act and similar regulations in other countries.

The involvement of companies like xAI is particularly noteworthy. Elon Musk has been both a vocal advocate for AI safety and a controversial figure due to his acquisition of Twitter (now X) and his outsized influence on public discourse. xAI’s Grok model has been positioned as a more “truth-seeking” alternative to ChatGPT, but it has also faced criticism for generating misleading content. Allowing government inspectors to see Grok’s code and training data before release could have significant implications for how the model behaves.

The Broader Implications

The disappearance of the CAISI page could undermine trust in the government’s commitment to transparency. Proponents of AI oversight argue that if the public cannot even rely on a press release to remain available, how can they trust that the evaluations are thorough? Critics of the agreements, on the other hand, might see the removal as a sign that the government is backtracking on promises to hold AI companies accountable.

Meanwhile, the global race for AI dominance continues. China has invested heavily in both AI development and regulation, often with less transparency. The U.S. government has repeatedly stressed the need to maintain a competitive edge while managing risks. The sudden disappearance of a page touting collaboration with industry leaders may signal internal dysfunction or a shift in strategy.

Gizmodo requested comment from the White House and the Commerce Department on Monday evening but did not receive an immediate response. The news organization will update its story if a reply comes. In the meantime, the only remaining record of the announcement is a static snapshot on the Wayback Machine, a reminder of how fragile digital government documents can be.

As of May 11, 2026, the URL redirects to the CAISI homepage, leaving no trace of the original text. For those following the evolution of AI policy, this disappearance is far more than a technical glitch—it is a data point in the ongoing story of how the United States navigates the promise and peril of artificial intelligence.

Source: Gizmodo News