Anthropic has presented the results of its Claude model's work, which, during analysis, identified critical vulnerabilities in two modern encryption systems. These are mathematical algorithms that currently protect logins, bank transactions, and private messages from prying eyes. Most notably, the machine discovered what leading global experts had missed.

Blow to HAWK: Key cracked in hours

HAWK is a promising digital signature protocol initially designed with the threat of quantum computers in mind. For two years, top specialists tested its strength and found no flaws. It took Claude just 60 hours to notice a hidden mathematical pattern. According to calculations, cracking the smallest HAWK key now requires 2^38 steps instead of the expected 2^64 — a reduction in computational cost by approximately 67 million times. There is a solution to the problem, but it is unpleasant: the key size will have to be doubled, negating HAWK's main advantage — compactness.

AES: Attack acceleration by hundreds of times

AES is the "gold standard" of encryption, which has protected the lion's share of internet traffic since 2001. Claude did not break full AES, but analyzed its simplified version (with 7 rounds instead of 10). The model found a non-standard approach, named "Möbius Bridge": one of the selection stages turned out to be redundant. As a result, the efficiency of the most powerful attack found was accelerated by 200–800 times. Claude needed three days to complete the entire task alone. Then, two specialists spent nearly a month double-checking its results.

Analyst's conclusions: The role of AI is growing

The main delay now occurs at the verification stage: humans can no longer keep up with the model's pace. This is reminiscent of the 2022 story, when another post-quantum algorithm, SIKE, was cracked in an hour on an ordinary laptop. These attacks pose no direct threat to Bitcoin and Ethereum — they operate on different principles. However, time is running out: Google plans to transition to new standards by 2029, and U.S. government agencies have a deadline of 2031. NIST now must evaluate a public attack on an algorithm that was still being considered for implementation. Anthropic has already indicated that this is not the last surprise — not all results have been published, and the company will soon hold a seminar with scientists.

My professional opinion: This case is a powerful signal for the entire industry. Artificial intelligence is ceasing to be just a tool for analysis and is becoming an independent "bug hunter." The cryptographic community must radically rethink approaches to security auditing: relying solely on the human factor is no longer possible. We need hybrid verification systems where AI and experts work in tandem, not sequentially.