Deep within Anthropic's glass-walled San Francisco headquarters, servers running the new Claude Mythos Preview model autonomously breached a 17-year-old remote code execution vulnerability in the FreeBSD operating system. The system navigated the server architecture without human intervention, isolating the flaw and exploiting it within seconds. By limiting access to this capability strictly to forty vetted corporate partners through Project Glasswing, Anthropic has effectively privatised the digital perimeter.

"The world needs to get ready," Anthropic co-founder Jack Clark told the Semafor World Economy summit on Monday, gripping the edge of his wooden podium. Clark confirmed that Mythos surpassed every internal benchmark for identifying vulnerabilities across legacy applications and operating systems. By gating this model away from independent researchers, the company enforces a cognitive enclosure of the security commons.

Independent security researchers are entirely priced out of the new defensive paradigm. Analysts at the startup AISLE successfully replicated the specific vulnerabilities Anthropic showcased, feeding the isolated code through smaller, commercially available models. However, the overarching capacity to autonomously audit millions of lines of code remains hoarded by institutions like Microsoft and JPMorgan Chase.

This structural hoarding represents the terminal financialisation of national security. The state and its proxy technology monopolies now extract sovereign rents by monopolising the tools required to secure digital infrastructure. A bifurcated landscape emerges wherein elite corporate entities operate heavily fortified digital fortresses, while secondary markets and municipal governments remain exposed to unpatched vulnerabilities.