The digital commons is being fenced off in real-time. Anthropic has announced the limited rollout of its 'Mythos' AI model, a tool capable of identifying and exploiting thousands of high-severity software vulnerabilities. Instead of releasing this capability to the public to harden global security, Anthropic is restricting access to a handpicked cabal of tech giants including Microsoft, Amazon, and Apple. This is 'Cognitive Enclosure' at its most literal.

By gating Mythos behind 'Project Glasswing,' the state and its corporate partners are establishing a technocratic monopoly over digital defense. While Anthropic claims this prevents 'bad actors' from utilizing the tool, it effectively ensures that only the elite possess the keys to the world’s software infrastructure. Leaked files suggest Mythos vastly surpasses public models in coding and reasoning, creating a tiered reality where the general public remains vulnerable to exploits that the corporate elite already have the power to fix.

This is the endgame of agentic AI. It is not a tool for liberation, but a mechanism for extracting security rents. As the US Treasury and Federal Reserve brief bank CEOs on the risks of 'automated exploit chaining,' the message is clear: safety is now a premium service. The professional class of software engineers faces deskilling, replaced by a black-box algorithm owned by the very firms that will profit from the resulting insecurity.