Claude solves long-standing cryptography mystery

Anthropic's AI model, Claude, solved a 25-year-old cryptography mystery in less than 30 minutes, according to a report by The-decoder. The mystery involved unexplained seeds used by the NSA to generate elliptic-curve digital signature parameters in 2000. Claude identified that these seeds originated from a Unix timestamp used with the lrand48 random number generator. Anthropic will give the $36,864 bounty to the Internet Security Research Group.

Why it matters

This result shows that large language models can perform deep pattern recognition that humans missed for decades. For a Swiss business, it means AI is moving from simple text generation to solving complex, technical problems. You should not view AI only as a tool for marketing or customer service. It can now act as a specialised auditor for your technical infrastructure and security protocols.

• AI can identify mathematical vulnerabilities in existing systems faster than manual audits.
• Security teams may need to integrate AI to scan for obscure patterns in proprietary code.
• The speed of discovery reduces the time you have to patch critical flaws.

What to watch

You should audit your proprietary algorithms for similar hidden patterns. If a model can identify a 25-year-old discrepancy in minutes, it can likely find flaws in your own code. Do not wait for a breach to test your security. Use AI to analyse your technical standards and ensure no legacy processes contain unintended weaknesses.

Within twelve months, automated mathematical audits will become standard. This requires AI models to move beyond pattern recognition and achieve verifiable precision in formal logic verification.