The Owner

The bottom line, above all

Artificial Intelligence Uncovers Ten Thousand Vulnerabilities as State Outsources Defense #

Sunday, 24 May 2026 · words

Close-up of a glowing server rack in a dark, climate-controlled data center, thick fiber optic cables routing through steel trays. 4K HDR professional photography, cool blue-grey colour palette, sharp lines, geometric precision, 50mm prime lens.
Close-up of a glowing server rack in a dark, climate-controlled data center, thick fiber optic cables routing through steel trays. 4K HDR professional photography, cool blue-grey colour palette, sharp lines, geometric precision, 50mm prime lens.

Inside a secure testing environment, Anthropic’s unreleased Claude Mythos Preview model systematically shredded the global software commons. In its first month of operation under Project Glasswing, the model autonomously discovered over 10,000 high- and critical-severity zero-day vulnerabilities across critical infrastructure. At Mozilla, the model was utilized to uncover and patch 271 vulnerabilities in Firefox 150—yielding ten times more findings than previous testing iterations. Biological triage is officially dead; human security analysts simply cannot read, process, or patch code at the speed of algorithmic discovery.

The Pentagon and intelligence agencies immediately recognized the sovereign failure of legacy cyber defense. The Department of Defense has established a task force to adopt and weaponize frontier AI models with hacking capabilities. Meanwhile, the executive branch capitulated to the technological elite. Hours before Donald Trump was set to sign a long-awaited executive order mandating a 90-day government safety review of new AI models, the president abruptly backed out, clearing a laissez-faire path for unchecked model deployment.

The judiciary is similarly paralyzed by the velocity of this cognitive enclosure. The D.C. Circuit Court of Appeals is currently reviewing the Pentagon’s March 3 designation of Anthropic as a supply chain risk. Judge Gregory Katsas, appointed by Trump, openly noted his "conceptual difficulties" with evaluating the software, stating that "AI three months from now will be totally different."

As the state retreats, enterprise capital must secure its own digital perimeter. Verizon has become the first telecommunications firm to join Project Glasswing. "Over the past several months, our information security team has been rigorously testing this critical new technology to determine its benefits to our network," said Verizon CEO Dan Schulman. When the federal government cannot regulate the code and cannot defend the grid, private balance sheets must absorb the cost of automated survival.