Anthropic's Mythos security model discovered previously unknown cryptographic weaknesses in a quantum-resistant algorithm under consideration for US standards, leading to its removal from contention. The finding demonstrates AI's role in identifying vulnerabilities in next-generation cryptography before standardization.
Anthropic's Claude Mythos Preview model discovered improved attacks against HAWK, a post-quantum digital signature scheme under NIST review, and round-reduced AES. The findings demonstrate AI language models' application to cryptanalysis.
Anthropic published research demonstrating how Claude, its AI assistant, can identify cryptographic weaknesses in security systems. The research highlights Claude's capabilities in detecting vulnerabilities that pose risks to encrypted communications.