The digital enclosure of advanced artificial intelligence is accelerating. Anthropic has abruptly halted the public release of its new Mythos model, citing its unprecedented ability to autonomously identify and exploit severe vulnerabilities across major operating systems. Instead of democratising this capability, Anthropic is restricting access to a 40-member consortium dubbed Project Glasswing.

This calculated restriction is not an act of altruism; it is a highly sophisticated rent-extraction maneuver. By monopolising the tool capable of exposing the accumulated incompetence of the biological developer class, Anthropic is forcing major enterprises into a capital-intensive security reliance. Firms like Apple, Microsoft, and AWS are now functionally dependent on Anthropic’s proprietary architecture to audit their own legacy codebases.

The recent 59.8-megabyte leak of Anthropic’s source code proved that human error remains the primary vector for systemic risk. Mythos represents the automated solution, capable of unearthing decades-old software flaws in seconds. For institutional investors, this signals a massive cybersecurity supercycle. The traditional software engineering paradigm is obsolete, and the firms hoarding the agentic models will dictate the pricing of digital survival.