Morning Minute — Tyler Warner
GM! Big story today: an AI model just exposed critical weaknesses in post-quantum cryptography.
Anthropic says an unreleased build of its most powerful model, Claude Mythos Preview, discovered two previously unknown attacks on cryptographic algorithms. One of those attacks hit HAWK — a digital-signature scheme designed specifically to resist quantum computers and previously advanced to round three of NIST’s post-quantum competition as the last lattice-based contender.
Why this matters: HAWK’s selling points were compact keys and fast signing, traits that matter a lot for blockchains because larger signatures consume block space and drive up fees. Claude found a mathematical symmetry in HAWK that human researchers hadn’t exploited, collapsing the work needed to break HAWK’s smallest key from 2^64 operations down to 2^38 — roughly a 67 million× reduction. Anthropic says the necessary fix essentially doubles key sizes, which “eliminates many of the reasons” HAWK was attractive in the first place.
To be clear: nothing in production is broken. HAWK hasn’t been deployed anywhere, and Bitcoin still runs on ECDSA (a pre-quantum scheme). But the discovery is a major development because the broader push to make crypto quantum-resistant — from institutional initiatives to roadmaps from Vitalik, NEAR, Zcash, and others — assumes the replacement schemes themselves are secure. Claude Mythos has just demonstrated that an advanced AI can find new, practical attacks on leading candidates before they ship.
There’s a silver lining: the same AI that finds vulnerabilities can be used to defend code. That’s why many teams are racing to point powerful models at their own cryptographic designs and implementations. The next question is now urgent: will AI break cryptography faster than it can help rebuild and harden it?
Other items: Corporate treasuries & ETFs, meme coin tracker — more in the newsletter.
Read more AI-generated news on: undefined/news
