Anthropic's AI model Mythos exposed secret US systems in just a few hours
The world of cybersecurity has received yet another confirmation that artificial intelligence is becoming not just a tool, but a key player in the field of national defense. The Mythos model, developed by Anthropic, demonstrated frightening efficiency by penetrating highly secure US government systems in a matter of hours, rather than the weeks that traditional methods would have required.
Details of Project Glasswing
The testing was conducted as part of a secret Anthropic initiative codenamed Project Glasswing. The goal was to identify critical vulnerabilities in the software that underpins national security, the economy, and public infrastructure. The results were staggering: Mythos found breaches in nearly every classified system it accessed, and in an unprecedentedly short amount of time.
Democratic Senator Mark Warner publicly mentioned these tests, citing data from a high-ranking source in the National Security Agency (NSA). He emphasized that the speed of the AI model fundamentally changes the rules of the game in cyber intelligence. Neither the NSA nor Anthropic representatives commented on the details, which only confirms the high level of secrecy surrounding the project.
Tension between innovation and security
This breakthrough comes amid growing tension between the US administration and technology giants. The Trump administration, concerned about potential risks, issued a directive restricting the use of Anthropic's latest models — Fable 5 and Mythos 5 — by foreign nationals. Moreover, strict controls were introduced on their public distribution.
Notably, over 100 leading cybersecurity experts, including representatives from Adobe and Nvidia, opposed these restrictions. In an open letter, they stated that depriving the country of access to the best AI auditing tools at a time when US adversaries are ramping up their capabilities is a dangerous and shortsighted strategy. They emphasize that Mythos is not unique in its kind, but its speed and efficiency make it an indispensable asset.
Cryptalist Expert Opinion: This incident is a clear marker that the AI arms race has moved into a practical phase. The speed with which Mythos hacked the systems proves that traditional cybersecurity methods are obsolete. The question is not whether to use AI for defense, but how to manage this weapon without losing the advantage. The market for AI-based cybersecurity solutions is poised for explosive growth, and investors who understand this first will come out ahead.