Key Takeaways
- Anthropic’s Claude Mythos broke NIST-backed HAWK-256 in about 60 hours, exposing a flaw that had eluded experts for two years.
- The attack does not affect Bitcoin, but underscores the growing role of AI in accelerating cryptographic research.
- The discovery could intensify scrutiny of Bitcoin’s planned post-quantum migration and other next-generation cryptographic standards.
Anthropic’s unreleased Claude Mythos Preview model has found a practical key-recovery attack on HAWK-256, a post-quantum digital signature candidate under active NIST evaluation, in roughly 60 hours of semi-autonomous work after the same flaw survived two years of expert human review.
The finding adds a new wrinkle as Bitcoin charts its quantum migration path.
HAWK carried real weight in the standardization race as the only lattice-based scheme among the nine candidates NIST advanced to the third round of its additional post-quantum signature process in May 2026.
Mythos uncovered a mathematical shortcut, known as a nontrivial automorphism, in the lattice structure underpinning HAWK’s security, cutting its effective key strength in half and forcing key sizes to double to maintain equivalent protection, a change Anthropic said would erase much of the scheme’s original appeal.
Cost figures underline the shift in cryptanalytic economics: Anthropic put the API bill at about $100,000, with a human researcher providing occasional project management guidance rather than lattice expertise.
Alongside HAWK, the model invented an attack technique it named the Möbius Bridge, making an existing attack on seven-round AES between 200 and 800 times faster.
Why Bitcoin Wallets Stay Safe
No deployed system is affected. Bitcoin uses ECDSA on secp256k1, which is entirely separate from HAWK, and HAWK has never been deployed in Bitcoin or any production system. Migration plans also remain intact, since BIP-360 targets ML-DSA and SLH-DSA, fully finalized NIST standards rather than third-round candidates.
Practical stakes for blockchains were nonetheless real: signature size consumes block space, block space translates into fees, and compact keys with fast signing formed HAWK’s entire pitch.
What the Break Signals
Speed comparisons cut in different directions. SIKE, another post-quantum candidate, survived years of expert review before being broken in roughly one hour on a laptop in 2022, so pre-deployment failures are precisely what the process exists to surface.
More striking is the verification gap: once Mythos returned the attack, Anthropic’s team spent several hundred hours confirming it was correct, meaning the AI found the flaw faster than experts could check its work.
Security analysts drew the uncomfortable inference. If AI-assisted cryptanalysis is being demonstrated publicly by safety-conscious labs, adversaries, including nation-state actors, are reasonably assumed to be pursuing similar capabilities behind closed doors.
Disclosure followed standard practice, as Anthropic shared both results with the algorithms’ authors, the US government, and industry partners before publishing, and coordinated the HAWK finding with NIST. Open questions now include whether the Bitcoin community accelerates formal AI-assisted cryptanalysis of ML-DSA and SLH-DSA as part of pre-activation review
Top Trending Crypto Articles
The information provided in this article is for informational purposes only. It is not intended to be, nor should it be construed as, financial advice. We do not make any warranties regarding the completeness, reliability, or accuracy of this information. All investments involve risk, and past performance does not guarantee future results. We recommend consulting a financial advisor before making any investment decisions.
Dr. Guneet Kaur is a senior editor at CCN.com and a Science Fellow at Exponential Science. She is a fintech and blockchain expert with extensive experience in digital finance education, blockchain ecosystems, and cryptocurrency markets. She has worked with global media such as Cointelegraph, as well as education and blockchain platforms, to design and lead strategic content and learning initiatives. As an educator and assessor for top-tier executive programs, she bridges real-world fintech trends with academic insight.
Dr. Kaur is also a published researcher and peer reviewer across fintech and data science journals, including Financial Innovation Journal and International Journal of Big Data Intelligence and Applications. Her work spans data-driven analysis, Web3 innovation, and technical content development. With a strong foundation in both industry and academia, she translates complex financial technologies into practical applications, empowering learners, professionals, and institutions across the rapidly evolving digital finance landscape.
