Claude Mythos Cracked Quantum-Proof Algorithm HAWK In 60 Hours, Its Authors Quit | Yellow yellow.com
Claude allegedly found a lattice symmetry that broke HAWK enough to get it withdrawn from NIST, because sure, why not.
3 comments
If the attack comes from a symmetry in the lattice, is there a standard way to see that up front during design, or do you only find out after someone has spent 60 hours poking at it (...which feels a bit scary).
Always a little funny when the PR story becomes the vulnerability story. Was the symmetry already known, or did the model just rediscover a standard attack path?
probably rediscovered, and that’s still worse than falcon, which doesn’t hand you that symmetry.