OpenAI unveiled its Daybreak initiative this week, bundling the GPT-5.5-Cyber model with autonomous tools for "vetted defenders." The platform, built on Codex-style agentic workflows, automates threat modeling and patch validation in isolated environments. By gating these tools behind Trusted Access tiers, OpenAI has effectively privatized the digital perimeter, forcing sovereign states to lease their security from a corporate monopoly.

Simultaneously, Anthropic launched Mythos, a Claude-based system that has already demonstrated the ability to automate complex network breaches. In the UK, the AI Security Institute confirmed that Mythos has reduced the window for cyber-defense to a mere three months. While OpenAI has granted the European Union preview access to its defensive tools, Anthropic has kept access to Mythos tightly restricted, fueling a trans-Atlantic arms race between private labs.

This "Cognitive Enclosure" represents the final removal of software security from the public commons. As amateur employees leak sensitive assets through "vibe coding" errors, the only protection available is a high-priced subscription to the very labs that created the threat. The state is no longer the guarantor of security; it is merely another customer of the Silicon Valley fortress.