The attempt to enclose the digital security commons has encountered its first structural failure. Anthropic Chief Executive Dario Amodei will walk into the White House on Friday to meet with Chief of Staff Susie Wiles. Their agenda centres on Anthropic's "Mythos" AI model—a system so proficient at exploiting zero-day vulnerabilities that the company refused to release it, instead gating it behind an elite consortium of banks known as Project Glasswing.

But the perimeter has already been compromised. "We’re investigating a report claiming unauthorised access to Claude Mythos Preview through one of our third-party vendor environments," an Anthropic spokesperson confirmed this week.

The macroeconomic lesson here is ruthless: sovereign security rents cannot be entirely protected by vendor contracts. The White House and financial giants like Goldman Sachs and Citigroup are desperate to hoard Mythos to establish private cyber-sovereignty. Yet, the vulnerability did not stem from the algorithm itself, but from the banal reality of enterprise IT supply chains. You can build the most advanced autonomous defence agent in human history, but if a third-party server administrator leaves a door open, the asset immediately becomes a public liability.