Anthropic's Claude AI uncovers HAWK, AES flaws in cryptography

Anthropic's AI, Claude Mythos, has demonstrated the capability to advance cryptography research significantly, as evidenced by two recent findings: a new attack on the HAWK digital signature scheme and a rapid attack on a reduced version of AES encryption. The HAWK attack found a previously unexploited symmetry, effectively halving the security of the scheme, while the AES attack applied a novel fingerprinting technique to improve attack efficiency.

The research showed that AI can achieve significant breakthroughs in cryptography, surpassing human efforts, and poses implications for academic research as it generates more results than can be validated by human experts.