TechCrunch reports that two cryptanalysts have used AI models from OpenAI and Anthropic to crack long-unsolved Enigma messages, a feat that echoes Alan Turing's wartime codebreaking work. Carter Leffen, a developer, asked OpenAI's Astra to search a database of Enigma messages, find an unbroken one, and decode it. Astra reportedly did its own archival research, found context clues, built an Enigma machine simulator, and recovered the plaintext of a message that had stumped researchers since 2005.
Frode Weirerud, who maintains the Crypto Cellar message database, validated Leffen's solution and said the model behaved like "a very professional cryptanalyst and archive researcher." He estimated that what Astra achieved in two days would take a human researcher weeks or months. Weirerud also noted that Astra's logs referenced archived messages in a "private collection" not hosted on his site, though it remains unclear whether the model accessed them or found material in German public archives.
A second cryptanalyst, Jack Willis, later told Weirerud he had used Anthropic's Claude Opus 5 to break a different unsolved message. Willis gave the model significantly more guidance, and it ultimately used the known signature of a particular officer's name to crack the code. Weirerud says just seven unbroken Enigma messages remain, along with one where the plaintext is known but the underlying code is still unsolved. The report is a single-source account, but it offers a concrete example of AI agents accelerating specialized historical and cryptanalytic work.