Claude Finds Crypto Flaws

Anthropic researchers have successfully utilized Claude to find mathematical weaknesses in cryptographic algorithms, including HAWK and a weaker version of AES. According to the research, the results do not have a practical impact on current computer systems. The study involved using Claude to identify potential flaws, with the model requiring specific prompting to explore possible vulnerabilities. The research demonstrates the potential of models like Claude in cryptographic analysis and the importance of carefully crafted prompts in eliciting meaningful responses. The full details of the research, including the prompts used, are available in the repository. As noted by Simon Willison, the research showcases Claude's capabilities in cryptographic research.