Two frontier AI models have independently completed a task that would have taken human cryptanalysts weeks or months: cracking World War II-era Enigma messages that had remained unsolved for decades.
Carter Leffen, a developer, simply told OpenAI's newest model, Astra, to search a database of Enigma messages for an unbroken message and decode it. The model did its own archival research, found context clues, built a simulator of the Enigma machine, and ultimately recovered the plaintext of a message that had baffled researchers since 2005. Frode Weierud, a retired electrical engineer who maintains the Crypto Cellar website with Enigma resources, validated Leffen's solution on September 18, saying he was left in "awe."
Notably, Astra's logs include discussion of archived messages in a "private collection" not hosted by Weierud — raising questions about how the model accessed external archives. "GPT-6 Astra is behaving like a very professional cryptanalyst and archive researcher," Weierud wrote. "What it has achieved in two days would take a human researcher weeks or even months."
On September 21, cybersecurity executive Jack Willis reached out to Weierud saying he had used Anthropic's Claude Opus 5 to break a different unsolved message, though he provided significantly more guidance to the model. Claude ultimately used the known signature of a particular officer's name to crack the code.
There are just seven unbroken Enigma messages remaining, along with one message where the plaintext is known but the code is still unbroken. With frontier AI models now tackling these puzzles, the last remnants of WWII's most famous cryptographic challenge may not survive much longer.

