History's long-secret codes are being cracked by amateurs using AI (www.nature.com)

Amateurs are leveraging advanced AI, particularly OpenAI's GPT-6 Astra, to crack long-secret codes, including a recently deciphered encrypted message from the German army that had remained unsolved for 85 years. Carter Leffen, a product development coach, utilized the LLM's capabilities to simulate an Enigma machine, decoding a message by testing billions of settings. This represents a remarkable intersection of AI technology and historical cryptography, allowing non-experts to successfully tackle complex problems that typically require extensive knowledge and experience. The implications for the AI/ML community are profound. As enthusiasts engage with large language models for historical codebreaking, they push the boundaries of AI applications beyond traditional domains. While this surge in interest is exciting, it also raises concerns among cryptography specialists due to the influx of potentially erroneous decryption claims. Experts emphasize the importance of validation amidst this new era of AI-enabled cryptanalysis, highlighting both the promise and pitfalls of machine-assisted codebreaking.