Anthropic Shows Claude Can Discover Mathematical Flaws in Cryptographic Algorithms
moyix · x · 2026-07-29
An Anthropic blog post details Claude's advanced capabilities in the field of cryptography.
While Claude previously demonstrated the ability to find implementation flaws in cryptographic code, researchers have now shown it can identify mathematical flaws in the algorithms themselves. Key breakthroughs include:
- Significantly weakening HAWK, a digital signature scheme designed for a post-quantum world.
- Discovering a new way to attack round-reduced AES, the most widely used symmetric cipher.
Although these findings do not currently affect production systems, they represent substantial research advances and highlight the profound implications of powerful AI models in reshaping digital security.
Related event: Claude Discovers New Weaknesses in Post-Quantum and AES Cryptography(9 posts)→
More from Safety
- User Argues: If You Hate Anthropic for Training Data, Hate Disney 100x More for Copyright Extension — jessi_cata · 2026-07-29
- OpenAI, Anthropic, Meta and DeepMind chiefs appear on the frontier slowdown letter — ShakeelHashim · 2026-07-29
- Another frontier AI employee says the slowdown letter reflects deep concern about RSI risks — morqon · 2026-07-29
- 1,122 frontier AI employees urge international tools to slow automated AI progress — scaling01 · 2026-07-29
- Pacing the Frontier publishes a 1,122-signature letter calling for AI slowdown tools — scaling01 · 2026-07-29
- Chip Restrictions Will Spur Chinese AI Innovation, Not Kill It — dbreunig · 2026-07-29