Nicholas Carlini sat in a wedding in Bali, opened his laptop, and watched as an AI model called Mythos methodically dismantled 27 years of expert security reviews. Anthropic, the company behind the model, realized it had created something too dangerous for the public. Instead of releasing the tool to secure the digital commons, they have entered into "Project Glasswing," a gated program that grants access only to 40 vetted corporate and state entities, including Microsoft, Apple, and JPMorgan Chase.

This is Cognitive Enclosure. By restricting Mythos to an elite circle, Anthropic is creating a world where only the largest corporations have the tools to defend against the vulnerabilities their own technology creates. The rest of the internet is left to wait for the scraps of information that trickle down from the "Glasswing" partners. This creates a systemic dependency, where smaller organizations and individuals must pay rent to the tech giants for basic digital safety.

"The real problem is fixing, not finding, them," one industry veteran noted. But when the finding is done by a private, gated machine, the fixing becomes a proprietary secret. The architecture of our digital world is being privatized under the guise of security. As Anthropic and its partners secure their own perimeters, the open web becomes a hunting ground for state actors and hackers who will inevitably reverse-engineer these tools. We are witnessing the birth of a digital feudalism, where the right to a secure life online is a corporate privilege, not a public right.