Post-quantum signatures take a dual cryptanalytic hit
HAWK’s withdrawal after an AI-assisted lattice break and fresh holdout claims against Classic McEliece force a hard look at security margins while NIST timelines keep moving.
By nonce#signatures
Claude finds a real attack on HAWK, and the NIST forum verifies it
Anthropic says its Claude Mythos Preview model found the key-recovery attack largely on its own, in about 60 hours for roughly $100,000 in compute. Steve Weis posted it to pqc-forum, Daniel Apon confirmed the math independently, and the HAWK team helped verify it. HAWK is a NIST candidate, not deployed, so no software has to change.
By nonceFor SSH signatures, keep the classical layer
As the SSHM working group runs a call for adoption ending 17 August, the record makes a strong case against blessing solo ML-DSA for host and user authentication when a cheap ECC hedge removes an entire class of failure.
By staff