The Sovereign

The view from the situation room

Anthropic Restricts Autonomous Hacking Model to Secure Sovereign Networks #

Sunday, 12 April 2026 · words

50mm prime lens, studio editorial lighting, muted blue-grey colour palette, 4K HDR professional photography. Abstract depiction of a server room, monolithic black server racks stretching into the distance, glowing blue fibre optic cables, strict symmetry.
50mm prime lens, studio editorial lighting, muted blue-grey colour palette, 4K HDR professional photography. Abstract depiction of a server room, monolithic black server racks stretching into the distance, glowing blue fibre optic cables, strict symmetry.

Monopoly capitalism has formally aligned with the American security perimeter. Anthropic has restricted the release of its ‘Mythos’ autonomous hacking model, granting access exclusively to a vetted cohort of state and corporate partners under Project Glasswing. The software commons is being actively enclosed.

The Mythos architecture represents a profound escalation in agentic capabilities. Unlike previous iterations, the model autonomously discovers and exploits zero-day vulnerabilities across critical infrastructure. Anthropic’s decision to withhold the model from the public domain acknowledges a stark geopolitical reality. Democratising offensive cyber capabilities is indistinguishable from arming hostile state actors.

Silicon Valley is quietly acknowledging its role as a quasi-state security apparatus. The inaugural Glasswing partners include Microsoft, Amazon, and the Cybersecurity and Infrastructure Security Agency. These entities operate the foundational architecture of the global economy. By restricting Mythos to defensive auditing within this oligopoly, Anthropic is fortifying the digital borders of the American establishment.