The race to build quantum-resistant cryptography suffered a major setback this week. A leading algorithm candidate known as HAWK was pulled from NIST's evaluation process after Anthropic's security model Mythos uncovered a critical flaw. The developer of HAWK announced the withdrawal Tuesday, effectively ending the algorithm's chance at becoming a US standard.
The HAWK Algorithm and Its Promise
HAWK was one of several candidates competing to become the official US standard for post-quantum cryptography. The National Institute of Standards and Technology runs a rigorous testing process to evaluate algorithms for their ability to withstand attacks from quantum computers. HAWK had survived two rounds of that process. Its withdrawal marks a significant loss for the standardization timeline.
How Mythos Found the Flaw
Anthropic, the AI company behind the Claude model, developed Mythos as a specialized security model. Mythos analyzed HAWK's mathematical structure and identified a vulnerability that rendered the algorithm broken. The finding was announced Monday. The developer withdrew HAWK from consideration the following day after confirming the issue.
This is not the first time AI has been used in cryptography, but it marks a significant milestone. An AI model uncovered a flaw in a formal NIST standardization process. The event underscores the dual role of AI as both a tool for security and a potential threat to established systems.
Why This Matters
The failure of HAWK has immediate consequences for the cybersecurity community. NIST's timeline for a post-quantum standard will likely face delays as other candidates are re-evaluated. Organizations planning their migration to quantum-resistant cryptography may need to adjust their roadmaps. The event also raises questions about the reliability of current PQC algorithms. If an AI model can find a flaw in a third-round candidate, other algorithms may harbor similar vulnerabilities. The broader implication is clear: the path to quantum-safe cryptography is longer and more uncertain than many believed. Companies and governments must now account for the possibility that AI-driven analysis will uncover more weaknesses, potentially reshaping the entire standardization process.



