Claude found mathematical flaws in two cryptographic algorithms that years of expert review missed

What happened

Anthropic’s Claude Mythos Preview AI discovered critical mathematical weaknesses in two cryptographic algorithms missed by years of expert analysis. The first flaw cuts the effective key strength of HAWK in half. HAWK is a post-quantum digital signature scheme currently under review by NIST. The second finding accelerates the best-known attack on seven-round AES, improving it 200 to 800 times over previous methods. AES is the most widely used symmetric encryption standard globally.

Why it matters

Cryptography is foundational for secure digital communication, financial transactions, and data privacy. HAWK’s vulnerability weakens a candidate for protecting systems against future quantum computer threats, raising risks for organizations planning their migration paths to post-quantum security. Cutting key strength in half effectively halves the security margin, forcing reconsideration or redesign.

The accelerated attack on seven-round AES signals that widely trusted encryption may be less robust than assumed against some advanced cryptanalysis. While full AES encryption uses more rounds, any weakening of partial-round variants shifts the risk calculus for hardware implementations, embedded devices, or edge systems that rely on reduced rounds for efficiency.

Both findings expose gaps in the cryptanalysis review process, showing that AI tools can spot subtle mathematical flaws that human experts overlook. This shifts incentives for security auditing and algorithm vetting toward integrating AI-driven analysis to tighten trust in cryptographic standards.

What to watch next

NIST’s ongoing post-quantum cryptography evaluation will likely face increased pressure to revisit HAWK and assess whether it still meets security goals. Developers and security architects should track updates on HAWK and adjust roadmap plans accordingly.

The cryptography community will need to evaluate how AI tools like Claude Mythos Preview reshape vulnerability discovery and whether these methods become standard practice. Watch for broader adoption of AI-assisted cryptanalysis, which could speed identification of weaknesses but also raise challenges around interpretability and trust in automated findings.

Enterprises relying on AES or post-quantum schemes should monitor official guidance and consider layered security approaches until these new vulnerabilities are fully understood and addressed.

AI Quick Briefs Editorial Desk