Cryptographic analysis has reached a new level. The Claude artificial intelligence model from Anthropic has discovered hidden weaknesses in two encryption systems that had eluded leading experts for a long time. These are mathematical algorithms that protect messages, logins, and banking data from outsiders.

It is important to emphasize: the current encryption standards used on the internet and in cryptocurrencies are not yet under threat. However, the very fact that a machine found what humans had missed over years of research changes the rules of the game in the field of cybersecurity.

Strike against the post-quantum signature HAWK

The first vulnerability was discovered in the HAWK digital signature system. This protocol was designed to be resistant to attacks from quantum computers. For two years, experts thoroughly checked its mathematical foundation and found no flaws. Claude needed only 60 hours to identify a hidden pattern.

According to my team's calculations, cracking the smallest HAWK key is now possible in 2^38 steps instead of the expected 2^64. This reduces computational costs by approximately 67 million times. A solution to the problem exists, but it is painful — the key size must be doubled, which negates HAWK's main performance advantage.

Attack on AES: the "Möbius Bridge" method

The second object of analysis was a simplified version of AES (Advanced Encryption Standard), the main encryption standard for internet traffic since 2001. Claude did not crack full AES but analyzed a version with 7 rounds instead of the standard 10. The result is impressive: the AI found a non-standard attack method, dubbed the "Möbius Bridge," which speeds up the search by 200–800 times.

Working alone, Claude took three days. Then two specialists spent nearly a month double-checking the result. This is a key point: humans can no longer keep up with the model's pace.

My expert opinion

This is not just a technical discovery, but a wake-up call for the entire industry. We are witnessing AI begin to perform functions that were previously the exclusive prerogative of humans — deep analysis and the search for hidden patterns. The question now is not whether AI can crack cryptography, but when it will find a breach in current standards. NIST and Google are already setting migration deadlines for post-quantum standards by 2029–2031, but discoveries like Anthropic's show that the timeline may be tighter than we think. Should the security of your data be entrusted to artificial intelligence? The answer is becoming increasingly clear.