Dave McGinnis watched the screen as Anthropic’s 'Mythos' model unraveled a 27-year-old security flaw in minutes. Instead of releasing this tool to the public to harden global systems, Anthropic has locked it behind Project Glasswing, a gated program for just 12 vetted corporate partners. This move marks the birth of Cognitive Enclosure, where the tools of digital survival are held as proprietary secrets.

The emergence of autonomous hacking models has triggered a global cybersecurity auditing supercycle that only the wealthiest firms can afford. While OpenAI responds by opening its GPT-Cyber model to more 'verified' defenders, the split is clear: safety is now a premium service. For the average user and small non-profit, the digital perimeter is becoming a gated community they cannot enter.

“The calculus around legacy systems has changed materially and immediately,” McGinnis stated during an IBM briefing. By gating the Mythos model, Anthropic is essentially extracting security rents from a terrified public. They have created a fire and are now charging for the only hose.

This privatization of the software commons mirrors the hollowing out of physical public utilities. As 'vibe coding'—building software via natural language—allows non-experts to flood app stores with unvetted code, the need for these elite AI guardians grows. We are witnessing the intentional deskilling of the professional class, replaced by a synthetic serfdom that serves the owners of the models.