Chris Fall looks at a screen that no longer exists. The Director of the Center for AI Standards and Innovation (CAISI) recently emphasized the need for "independent, rigorous measurement science," according to Politico and NIST notices. By Monday afternoon, however, the U.S. Commerce Department website returned a stark error for its AI vetting deals: "Sorry, we cannot find that page." This digital scrubbing erases public details of security agreements with Google DeepMind, Microsoft, and xAI, according to reporting by Reuters. The government had originally pledged to conduct "pre-deployment evaluations and targeted research" on unreleased models.

As the state blinds itself to the internal workings of frontier models, the "Cognitive Enclosure" is accelerating in the private sector. On May 6, xAI—recently rebranded as SpaceXAI—signed a massive agreement giving Anthropic exclusive access to the Colossus 1 supercomputer in Memphis, per CryptoBriefing. This facility is a physical fortress of intelligence, housing 220,000 NVIDIA GPUs and consuming 300 megawatts of power. The deal is expected to generate between $3 billion and $6 billion in annual revenue for xAI, according to New Street Research and Forbes.

The thread linking these, though stated in no filing, is the rapid privatization of the digital perimeter. As oversight pages vanish into 404 errors, the physical hardware of intelligence is seining into billionaire hands; the causal link between the regulatory blackout and the $6 billion compute deal remains unconfirmed by this paper’s sources. What is clear is that the public’s right to know the dangers of these models has been traded for corporate revenue streams. We are witnessing the birth of a world where the state is a mere spectator to the algorithms that will govern us.