Anthropic says Claude found new attacks on HAWK and reduced-round AES
matthew_d_green · x · 2026-07-29
Anthropic says Claude Mythos Preview helped researchers discover new ways to attack cryptographic algorithms.
- The first result weakens HAWK, a post-quantum digital signature scheme.
- The second finds a new attack on round-reduced AES, the most widely used symmetric cipher.
- Anthropic says these are substantial research advances, but do not affect production systems today.
- The post argues that powerful AI models may increasingly help uncover flaws in cryptography itself, not just implementation bugs in code.
Related event: Anthropic's Claude Mythos Uncovers Flaws in Post-Quantum Crypto and AES(31 posts)→
More from AGI Musings
- A X debate argues ASI may not need a “superintelligent internet” of training data — ___Patrice___ · 2026-07-29
- A Gemini quote predicts society’s AI phase change will arrive between late 2027 and mid-2029 — MattGarciaEth · 2026-07-29
- Matt Green: verifying flashy model-generated math results is now the hard part — matthew_d_green · 2026-07-29
- AI companies may be close to automating AI research, with control risks rising — FlorianGallwitz · 2026-07-29
- A calculator-style analogy argues AI produces representations, not thoughts — gerardsans · 2026-07-29
- Matthew D. Green says advanced AI could leave everyone in the same boat — matthew_d_green · 2026-07-29