Anthropic has presented the results of a sensational study: its AI model Claude discovered hidden mathematical vulnerabilities in two cryptographic systems. These are fundamental algorithms that currently protect billions of messages, logins, and bank transactions from unauthorized access.
It is important to emphasize: the current encryption systems used in everyday life are not yet under threat. However, the very fact that a machine found in a matter of days what leading global experts had missed over two years changes the rules of the game in the field of cybersecurity.
HAWK Vulnerability: 67 Million Times Faster
HAWK is a promising digital signature system designed with the threat of quantum computers in mind. For two years, specialists could not find any flaws in it. But Claude needed only 60 hours to identify a hidden pattern in the mathematical foundation of HAWK.
According to my analysis of Anthropic's data, cracking the smallest HAWK key is now possible in 2^38 steps instead of 2^64. This means a reduction in computational costs by approximately 67 million times. There is a solution to the problem, but it is painful: it will require doubling the key size, which completely negates HAWK's main advantage—compactness.
AES and the "Möbius Bridge" Method
AES (Advanced Encryption Standard) is the gold standard of encryption, which has been protecting most internet traffic since 2001. Claude could not crack the full 10-round AES, but it analyzed a simplified version with 7 rounds.
The AI found a non-standard attack method, called the "Möbius Bridge." The essence is that one of the key selection stages turned out to be redundant. As a result, the most effective attack was accelerated by 200–800 times. Notably, Claude worked alone for three days, and then two specialists spent nearly a month double-checking its results.
Implications for the Crypto Industry
The main problem now is not the speed of AI, but the delay at the verification stage. Humans physically cannot keep up with the model's pace. This is not the first case—in 2022, researchers cracked another post-quantum algorithm, SIKE, in an hour on a regular laptop, but then humans found the error.
Bitcoin (BTC) and Ethereum (ETH) operate on different principles, and these attacks do not affect them. However, time is running: Google plans to switch to new standards by 2029, and U.S. government agencies have a quantum deadline of 2031.
Now NIST will have to evaluate a public attack on an algorithm that was still being considered for implementation. Anthropic has indicated that not all results have been published. In the near future, the company will hold a closed seminar with leading scientists.
My expert assessment: This event is a clear signal for the entire crypto industry. AI is no longer just an analysis tool, but an active participant in security audits. Trusting data protection verification solely to humans is becoming not just inefficient, but potentially dangerous. The question is not whether AI will replace cryptographers, but how quickly we can adapt verification processes to the new reality.