AIWHYS

Technology

cryptanalysis

2 synchronized stories carry this tag.

ReliabilityResearch release

Claude Mythos helped find two new cryptanalytic improvements

Anthropic researchers used Claude Mythos Preview to improve key recovery against the HAWK-256 challenge parameter and accelerate an attack on seven-round AES-128 by 200–800 times. Papers and HAWK demonstration code were released.

Why it matters: AI-assisted research can now produce technically checkable advances in a high-stakes field, making expert verification capacity a limiting resource.

Read storyOriginal source · anthropic.com
ReliabilityPreprint

CryptanalysisBench tests frontier models on verifiable cipher attacks

CryptanalysisBench packages 191 tasks across six cryptographic primitive families and three difficulty tiers. Five frontier models solved 65–86% of known-break tasks and reportedly found previously unknown issues in SpoC and KINDI.

Why it matters: Cryptanalysis is a rare reasoning domain where candidate outputs can often be checked mechanically, making it useful for measuring both capability and emerging security risk.

Read storyOriginal source · arxiv.org