Anthropic: Claude Discovers Fatal Flaw in HAWK, Breaks NIST Post-Quantum Signature Candidate Algorithm in Just 60 Hours
Anthropic's Claude Mythos Preview model discovered a fatal flaw in the NIST post-quantum signature candidate algorithm HAWK within 60 hours, leading to the algorithm's withdrawal.
Anthropic demonstrated the research capabilities of its Claude Mythos Preview model in the field of complex cryptography. The model not only analyzed the NIST post-quantum signature candidate algorithm HAWK but also proposed more efficient attack methods against a simplified version of the AES-128 encryption standard.
This breakthrough indicates that large language models can now deeply participate in high-difficulty scientific discovery processes. Cryptographic vulnerability mining that traditionally required experts to spend months or even years was completed by the AI model in 60 hours, significantly improving research efficiency.
The HAWK algorithm was therefore withdrawn by its developers from the NIST standardization process, highlighting the actual impact of AI in cybersecurity assessment. In the future, AI may become an important auxiliary tool for cryptographic standard setting and security verification, but it also brings new security challenges.
This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.