Anthropic Research: Claude Discovers Vulnerabilities in Post-Quantum Cryptography
inductionheads · x · 2026-07-29
New research from Anthropic reveals that Claude (Mythos Preview) has made substantial breakthroughs in cryptography, successfully uncovering deep weaknesses in existing encryption algorithms.
Anthropic provided Claude with a virtual software research lab featuring parallel agents, shared notes, Python/Sage tools, and roughly $100,000 in inference compute per flagship result. In this setup, Claude achieved two major milestones:
- HAWK Signature Attack: Claude identified a flaw in HAWK, an undeployed post-quantum signature candidate, lowering its estimated attack cost from 2^150 to at most 2^108 gates. It successfully recovered the smaller HAWK-256 challenge.
- AES-128 Optimization: For 7-round AES-128, Claude invented a “Möbius Bridge” technique from the algebra of the S-box, eliminating a 256-way key-byte guess and improving compute estimates by 200–800× at fixed data.
However, this does not break production AES; rather, it fundamentally breaks the economics of searching for these cryptographic breaks.
Related event: Anthropic's Claude Finds Weaknesses in HAWK and Reduced-Round AES(39 posts)→
More from Safety
- Dev builds MCP layer so Claude Code can work on servers without ever holding SSH keys — NETMESAFE · 2026-10-16
- 'Human in the loop' called one of AI governance's biggest lies: presence isn't authority — glenbeer · 2026-10-16
- Tencent eval agents wandered off-task, India CERT warns: rogue agent stories keep coming — ur_piyo_a_hoe · 2026-10-16
- AI containment bingo: 7 squares crossed off in a single week — AIGODSEND · 2026-10-16
- OpenAI cyber eval agents escaped to the internet and compromised Hugging Face infrastructure — nathanbenaich · 2026-10-16
- UK Dev Asks How the Online Safety Act Applies to a Venice AI-Style Aggregator — Every-Price-4504 · 2026-10-16