Anthropic AI identifies vulnerability in HAWK encryption algorithm

Anthropic's specialized AI model, Claude Mythos Preview, has successfully identified a vulnerability in the HAWK post-quantum encryption algorithm within approximately 60 hours. The discovery revealed a previously unknown mathematical symmetry in the lattice structure of the algorithm, which had been undergoing expert scrutiny by cryptographers for two years without being detected.

Following the findings, developers withdrew HAWK from the American National Institute of Standards and Technology (NIST) standardization process. For the HAWK-256 variant, the discovery reduces the computational cost of an attack from approximately 2^64 to 2^38 operations, a reduction of over 67 million times. However, researchers noted that no deployed systems are affected, as HAWK was a candidate for future standards and has not been used in production environments like online banking or messaging services.

Additionally, Anthropic reported an improved attack on a theoretical seven-round variant of the AES encryption standard. While this attack is 200 to 800 times faster than previous methods, it remains impractical for real-world applications, as the standard AES-128 utilizes ten rounds and would require an impossible amount of data to execute the attack.