Anthropic Says Claude Mythos Found HAWK, 7-Round AES Flaws
TL;DR
- HAWK was withdrawn from NIST PQC standardization the same day Anthropic published, marking the fastest algorithm retirement in modern post-quantum review.
- Human verification consumed several hundred researcher hours over nearly a month, making validation the binding bottleneck in AI-accelerated cryptanalysis.
- The AES finding targets a flaw in the algorithm's mathematics rather than its implementation, a qualitatively different threat class from prior AI security work.
A single Anthropic researcher, working with an unreleased model called Claude Mythos Preview for roughly 60 hours, reportedly found a mathematical shortcut against HAWK, one of the post-quantum digital signature candidates still in NIST's standardization process. The find is not an implementation bug. It is a previously unexploited symmetry in HAWK's lattice structure, and it drops the expected attack cost on HAWK-256 from roughly 2^64 operations to 2^38. That is close to a practical break for the small-parameter version. HAWK is not deployed anywhere, which is the honest reason nobody has to patch anything today. Anthropic also released a Zeta-function bound improvement from another unreleased Claude today.
The second finding is more of a stress test than a scare. Given about 3 days of autonomous work, on the order of a billion output tokens, and another six figures in API spend, the same model produced a technique it calls the Möbius Bridge, which speeds up the best known attack on 7-round AES-128 by 200 to 800 times depending on how you measure it. Full AES has 10 rounds, and the 10-round cipher remains unbroken. Round-reduced attacks are how cryptographers probe cipher margins, so this belongs on the research pile, not in a CVE.
What makes the write-up interesting to non-cryptographers is the economics. Two publishable cryptanalysis results, each for about $100,000 in inference, on schemes that had already been picked over by human experts. Anthropic notes that a lot of the wall-clock time was researchers validating the model's output, not the model doing the math, and that Mythos needed a 'good amount of prompting' before it stopped concluding the problems were impossible. Preliminary results are mentioned on LEA, Serpent-128, Salsa20, Poseidon, and SHA-1, some more consequential than others.
A single vendor writing up its own unreleased model is not the same as independent replication. Anthropic says it consulted academics, shared advance copies with US government and industry partners, and coordinated public disclosure with HAWK's authors on the NIST mailing list in June, but the write-up does not name the outside cryptographers who checked the work or say whether HAWK's team accepts the analysis. Until outside cryptographers publish their own view, the specific numbers stand as Anthropic's alone.
If this line of work generalizes, the winners are standards bodies and internal security teams who now have a plausible way to run AI-assisted cryptanalysis over borderline candidates before they get baked into products. The bottleneck moves from finding attacks to validating them, which is a much better problem to have than the one we had last week.
What others are reporting
-
The New York Times Read →
Filed under Politics rather than Technology, signaling the findings have entered national-security policy discussion beyond the security research community.
-
CyberScoop Read →
Reports Anthropic coordinated with NIST and government partners ahead of publication; includes enterprise expert commentary framing cryptography readiness as ongoing operational work.
The attacks described in these two papers are the strongest attacks we have found to date.
-
The Hacker News Read →
Deepest technical coverage; confirms HAWK was withdrawn from standardization the day of publication and documents the researcher prompts that pushed Mythos past its initial refusals.
-
The Next Web Read →
Frames findings as a leap from implementation bugs to algorithmic mathematics; raises the absence of any government mechanism to coordinate AI-discovered flaws in deployed production systems.
In just one year, language models have gone from being unable to perform cryptanalysis to finding flaws.
-
Decrypt Read →
Leads on the human verification bottleneck and quotes Anthropic's concern that human researchers may become the constraint in AI-accelerated security triage.
Claude found a symmetry buried in HAWK's math that no human had thought to use. Key recovery cost fell from 2^64 operations to 2^38.
-
CSO Online Read →
Details the research scaffold Anthropic built for autonomous hypothesis generation and frames the month-long verification effort as the new rate-limiter in the discovery pipeline.
Despite Hawk having survived two rounds of expert human review over a period of two years, Mythos was able to improve the best-known attack on it in just 60 hours of work.
Shared on Bluesky by 5 AI experts
-
this will take a while to diffuse, but the idea that LLM results derive primarily from the expertise of the operator is fully dead www.anthropic.com/research/dis...
View on Bluesky →
Originally reported by anthropic.com
Read the original article →Original headline: Anthropic Says Claude Mythos Preview Discovered New Attacks on HAWK Post-Quantum Signature and 7-Round AES