Anthropic's Mythos AI breaks HAWK quantum-resistant algorithm
Anthropic's Mythos AI security model has uncovered a critical flaw in HAWK, a post-quantum cryptography algorithm under consideration as a US standard by NIST. The flaw, found during the third round of testing, rendered HAWK broken, leading its developer to withdraw the algorithm on Tuesday, July 21, 2026. Mythos also found weaknesses in the widely used AES cipher. The results are incremental and do not break current cryptosystems, but they signal potential advances in breaking cryptography crucial to privacy and security. The tests used weakened versions of the algorithms, as is standard for adversarial peer review.
Key facts
- Anthropic's Mythos AI security model found a flaw in HAWK, a post-quantum cryptography algorithm.
- HAWK was under consideration as an official US standard by NIST.
- HAWK had survived two rounds of testing and was in a third round when the flaw was discovered.
- The developer of HAWK withdrew the algorithm on Tuesday, July 21, 2026.
- Mythos also found weaknesses in the widely used AES cipher.
- The results are incremental and do not break any cryptosystems currently in use.
- Tests used weakened versions of the algorithms, as is standard for adversarial peer review.
- The findings could signal important advances in breaking cryptography.
Entities
Institutions
- Anthropic
- NIST (National Institute of Standards and Technology)