Anthropic has presented sensational results: its language model Claude discovered critical weaknesses in two encryption systems that were considered reliable. These are mathematical algorithms that protect messages, logins, and banking data from outsiders. While current standards like AES have not been affected so far, the very fact that a machine found what leading cryptographers missed changes the rules of the game.

What is HAWK and what vulnerability was found

HAWK is a digital signature designed with the threat of quantum computers in mind. For two years, experts could not find any flaws in it. It took Claude only 60 hours to notice a hidden pattern in the mathematical foundation of the algorithm. According to calculations, cracking the smallest HAWK key is now possible in 2^38 steps instead of 2^64, reducing computational costs by approximately 67 million times. There is a solution to the problem, but it is unpleasant: doubling the key size, which negates HAWK's main advantage.

What the AES test showed

AES is the gold standard of encryption, supporting most internet traffic since 2001. Claude did not crack full AES, but analyzed a simplified version with 7 rounds instead of the standard 10. The model developed a non-standard attack called Möbius Bridge, which speeds up key selection by 200–800 times. Claude spent three days working alone on the entire task, and then two specialists spent nearly a month double-checking the result.

What is the significance of this work

The main delay now occurs at the verification stage: humans simply 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. However, the key point here is that AI acts autonomously and finds what experts miss.

Bitcoin and Ethereum operate on a different principle, so these attacks do not affect them. But time is running: Google plans to transition to new standards by 2029, and U.S. government agencies have a deadline of 2031.

What to watch for next

Now NIST will have to evaluate the public attack on an algorithm that was still being considered for implementation. Anthropic indicated that not all attack results have been published yet — the company will soon hold a seminar with scientists. The question remains open: should the verification of your data security be entrusted to artificial intelligence?

Expert opinion: This is not just a technical novelty, but a signal that traditional security audit methods are becoming obsolete. AI is already capable of performing work that used to take teams years, and this calls many current standards into question. The crypto industry should prepare for a new era — where protection will be built taking into account the capabilities of machine analysis.