Code Cracked: Why the US Actually Blocked Anthropic’s Flagship Models

The main geopolitical intrigue of June has received a technical explanation. On June 24, 2026, the real trigger behind the US Commerce Department’s unprecedented order to shut down access to Anthropic models (the June 12-13 case) was revealed. During testing under `Project Glasswing`, the Mythos model successfully identified critical vulnerabilities in classified US government systems.

This incident has blurred the line between the civilian IT sector and cyber-weaponry. Frontier models have crossed a threshold where they outperform the Pentagon’s staff information security auditors in analyzing closed networks. The realization that access to such a "cracker" could be obtained via a standard cloud API caused panic in the White House, leading to an immediate export ban. The recent "amnesty" by Donald Trump now makes sense: the government simply cannot afford to destroy such a powerful analytical tool, so it is being placed under strict state supervision.

Source: Anthropic / AP / Reuters