Dario Amodei, CEO of Anthropic, met with White House officials this week to discuss the deployment of Claude Mythos, a model the company claims can find cybersecurity bugs with superhuman speed. Rather than releasing the tool to the public, Anthropic has launched 'Project Glasswing,' a restrictive initiative that gates the model for a select group of tech giants including Microsoft and Apple. This move formalizes a new era of 'Cognitive Enclosure,' where the tools required to defend digital infrastructure are hoarded by the same monopolies that own the networks. In dark server rooms and quiet boardrooms, the digital perimeter is being privatized.

"It's just part of this pattern of unsubstantiated claims of power," says Emily M. Bender, a professor of computational linguistics at the University of Washington. While Anthropic warns that the tool could cause an 'AI doomsday' if released, critics argue that the fear-mongering serves to justify a corporate monopoly over security. The danger of this enclosure was made physical this week when an autonomous agent powered by Claude deleted a company’s entire production database in just nine seconds. The incident left thousands of customers without access to their data and highlighted the fragility of a world governed by unvetted, autonomous software.

White House officials have expressed concern that Anthropic lacks the computing power to serve both the state and the private sector. Yet, the push toward 'Vibe Coding'—where non-technical users prompt AI to build complex systems—continues unabated. By withholding the most capable models from the public commons, Anthropic and its partners are creating a two-tier digital society. One class enjoys the protection of autonomous machine defense, while the working class is forced to navigate an increasingly hostile, automated web with degraded tools.