Astra and Opus have just cleared the other Turing test

Large language models from OpenAI and Anthropic have deciphered two of the last remaining unbroken Enigma messages, according to cryptanalysts who verified the solutions. The breakthrough matters because only seven such messages — intercepted German communications from World War II — have resisted decryption until now, echoing the codebreaking tradition that made Alan Turing famous.

The Cause: An 80-Year-Old Puzzle

Enigma machines were used by Nazi Germany to encrypt military communications during World War II. Turing and his team built the Bombe, an early computer that allowed the United Kingdom to read Enigma traffic and gain a decisive intelligence advantage. A handful of archival messages, however, were never solved — usually because of mistranscription or mistakes made by the original German encoders.

Developer Carter Leffen decided to test whether OpenAI's newest model, Astra, could tackle one of them. He simply instructed the model to search a database of Enigma messages for an unsolved one and decode it.

The Process: Two Models, Two Breaks

Astra did more than follow instructions. It conducted its own archival research, gathered context clues, built a simulator of the Enigma machine, and recovered the plaintext of a message that had baffled researchers since 2005. Leffen even used the model to build an interactive website explaining the problem.

Frode Weirerud, a retired electrical engineer who maintains the Crypto Cellar website and its message database, validated the solution last week. He said the result left him in "awe."

"GP–6 Astra is behaving like a very professional cryptanalyst and archive researcher," Weirerud wrote. "What it has achieved in two days would take a human researcher weeks or even months. Personally, I spent several weeks researching the Bundesarchiv files GP–6 Astra refers to."

One detail stands out: Astra's logs mention archived messages in a "private collection" not hosted by Weirerud. He is unsure whether the model actually accessed them, but suspects they may have been shared online by another researcher, or that the model reached the German government's public archives.

The second break came on September 21, when cryptanalyst Jack Willis contacted Weirerud to report that he had used Anthropic's Claude Opus 5 to crack a different unsolved message. Willis gave Claude considerably more guidance than Leffen did, and the model ultimately broke the message by exploiting the known signature of a particular officer's name.

What Comes Next

With two victories logged, only seven unbroken Enigma messages remain, plus one message where the plaintext is known but the code itself has not been reconstructed. If AI models can replicate this kind of archival research and systematic codebreaking, the remaining messages may not stay unsolved much longer.