Researchers use OpenAI Astra and Claude Opus 5 to decode Enigma messages

In short: Two researchers say they used AI models from OpenAI and Anthropic to decode two Enigma-coded messages from World War II that had stayed unsolved for years.

What happened

Computer scientist Alan Turing helped crack Nazi Germany’s Enigma code during World War II. Enigma was a code machine that worked like a letter-scrambling typewriter, and it was used to send military messages.

TechCrunch reports that some Enigma messages stored in archives have remained unbroken, often because the original message was copied down incorrectly or had mistakes. Now, cryptanalysts say modern AI systems were able to solve two of these stubborn cases.

Developer Carter Leffen asked OpenAI’s newest model, called Astra, to find an unbroken message in a database and decode it. According to the report, the model did several steps on its own, including looking for historical context clues, building a software simulator of the Enigma machine, and recovering the “plaintext” (the normal, readable message before it was encoded). Frode Weirerud, who runs the Crypto Cellar website and maintains related records, said he validated Leffen’s solution.

Separately, another cryptanalyst, Jack Willis, told Weirerud he used Anthropic’s Claude Opus 5 to break a different unsolved Enigma message. In this case, Willis gave the model more guidance, including a known name “signature” that helped narrow down the answer.

Why it matters

This is a clear example of today’s AI being used as a research assistant, not just a chatbot. If these results hold up, historians could decode more archived messages faster, and the number of remaining unsolved Enigma messages could shrink from just a few to none.

Source: TechCrunch AI