The world of cryptography has encountered an unexpected discovery. Anthropic's artificial intelligence, Claude, has found critical vulnerabilities in two encryption systems that underpin the protection of messages, passwords, and banking data. This is not about current standards being broken—the threat turned out to be much deeper and more intriguing.

The machine found what leading experts had missed. This changes the rules of the game in security auditing.

Vulnerability in HAWK: 67 million times faster

The HAWK system is a digital signature designed with future quantum threats in mind. For two years, experts meticulously checked its mathematical foundation and found no flaws. However, Claude needed only 60 hours to uncover a hidden pattern.

According to my calculations, based on published data, cracking the smallest HAWK key is now possible in 2^38 steps instead of 2^64. This means computational costs have been reduced by approximately 67 million times. There is a solution, but it is unpleasant: the key size will have to be doubled, negating HAWK's main advantage—its efficiency.

Test on AES: "Möbius Bridge" and 800x acceleration

AES (Advanced Encryption Standard) is the backbone of the modern internet, in use since 2001. Claude did not crack the full 10-round AES, but analyzed its simplified version with 7 rounds. The AI found a non-standard approach, dubbed the "Möbius Bridge," which allowed skipping one of the selection stages.

Result: the most effective attack on this version accelerated by 200–800 times. It took Claude three days to complete the task. After that, two specialists needed nearly a month to double-check the result. This vividly demonstrates how much slower the human factor is.

Why this matters for the crypto industry

The main delay now occurs at the verification stage. Anthropic directly states: humans can no longer keep up with the model's pace. This has happened before—in 2022, researchers cracked the post-quantum algorithm SIKE on a regular laptop in an hour.

It is important to emphasize: Bitcoin (BTC) and Ethereum (ETH) operate on different principles, and these attacks do not affect them. However, time is running out. Google plans to transition to new standards by 2029, and U.S. government agencies have a quantum deadline of 2031.

What's next

NIST now must evaluate the public attack on an algorithm that was still being considered for implementation. Anthropic has indicated that not all results have been published. The company will soon hold a special seminar with scientists.

My conclusion as an analyst: this work is not just a technical curiosity. It calls into question the very methodology of security verification. If AI can find hidden patterns in mathematics within hours, then entrusting audits solely to humans is a risky luxury. The question is not whether we will be hacked, but whether we will have time to prepare.