Anthropic's artificial intelligence Claude has made a breakthrough in cryptanalysis, identifying previously unknown vulnerabilities in two encryption systems. These involve mathematical algorithms that underpin the protection of messages, logins, and banking data. This discovery does not threaten current security standards but demonstrates a new era in finding cryptographic flaws.

Vulnerability in HAWK: 67 million times faster

The HAWK system, developed as a quantum-resistant digital signature, was considered reliable after two years of expert review. However, Claude needed only 60 hours to uncover a hidden mathematical pattern at its core. According to Anthropic's analysis, cracking the smallest HAWK key is now possible in 2^38 steps instead of 2^64 — a reduction in computational cost by approximately 67 million times. Solving the problem requires doubling the key size, which effectively negates HAWK's main advantage: its efficiency.

Möbius Bridge attack on simplified AES

Regarding AES (Advanced Encryption Standard) — the primary encryption standard for internet traffic since 2001 — Claude was unable to crack the full version. Instead, the AI analyzed a simplified version of the algorithm with 7 rounds instead of the standard 10. The result was an attack called Möbius Bridge, which accelerated the most effective cracking method by 200-800 times. Notably, Claude worked alone for three days, and then two specialists spent nearly a month double-checking its findings.

Significance for the crypto industry

This case highlights a key problem: humans can no longer keep up with the pace of AI. The main delay now occurs at the stage of verifying results. This is not the first time post-quantum schemes have been broken — in 2022, the SIKE algorithm was cracked on an ordinary laptop in an hour. However, Bitcoin (BTC) and Ethereum (ETH) operate on different principles, and such attacks do not affect them. Nevertheless, time is running out: Google plans to transition to new standards by 2029, and U.S. government agencies have a deadline until 2031.

My expert conclusion: This case is not just a technical finding but a wake-up call for the entire crypto industry. AI is beginning to outpace human expertise in finding vulnerabilities, and we need to adapt security verification processes. NIST will need to assess this attack, and I expect Anthropic to hold a closed seminar with scientists in the near future. The question is not whether AI will break current standards, but how quickly we can implement new ones resistant to such methods.