Discovering Cryptographic Weaknesses

Shows a frontier model finding flaws in cryptographic algorithms themselves, not just implementation bugs, which changes assumptions about AI-driven security research.

• articleAlignment Faking
• articleOff Switch Dual Use
• articleAutomated Researchers Mitigate Alignment Failures
• articleMultiagent Systems

• articleDiscovering Cryptographic Weaknesses with ClaudeAnthropicAI
• articleNew Attacks on HAWK and AES Cryptographic SchemesAnthropicAI
• blogAre Mythos’ cyber capabilities overhyped?Timothée Chauvin

Checking sign-in…

Loading comments…