AI models successfully decipher long-unsolved enigma messages

Two AI models, Astra from OpenAI and Opus 5 from Anthropic, have successfully decrypted two long-unsolved Enigma messages, marking a significant breakthrough in historical cryptography. Carter Leffen, a developer, used Astra to decode an Enigma message that had puzzled researchers since 2005. The model performed archival research, identified context clues, built an Enigma simulator, and ultimately recovered the plaintext. Frode Weierud, a retired electrical engineer and cryptology enthusiast, validated Leffen’s solution, expressing awe at the model’s capabilities. Weierud noted that Astra’s achievement, which took just two days, would have taken a human researcher weeks or months. Meanwhile, Jack Willis used Opus 5 to break another message by leveraging known officer signatures. Weierud highlighted that only seven Enigma messages remain unsolved, with one having a known plaintext but unbroken code. The breakthrough underscores AI’s growing role in historical and cryptographic research. The models’ logs suggest they may have accessed private collections or public archives, though the exact sources remain unclear. Weierud emphasized the efficiency of AI in handling complex cryptographic tasks, noting that the models’ performance surpassed traditional methods. These achievements highlight the potential of large language models in solving historical puzzles and advancing cryptanalysis. As AI continues to evolve, its applications in decoding historical data and enhancing security are expected to expand significantly.