Alan Turing is widely remembered for the intelligence test that bears his name. But during World War II, his more consequential legacy was helping break the Enigma code used by Nazi Germany — a breakthrough that helped shift the course of history.
Now, cryptanalysts say large language models from OpenAI and Anthropic have solved two long-unbroken Enigma messages that had resisted decoding for decades.
Follow THE FUTURE on LinkedIn, Facebook, Instagram, X and Telegram
The Last Remaining Pieces Of A Wartime Puzzle
During the war, Turing and his colleagues built the Bombe, an early machine designed to help the United Kingdom decipher Enigma traffic. Yet a small number of archival messages remain unresolved today, often because of transcription mistakes or errors in the original enciphering process.
In one case, developer Carter Leffen asked OpenAI’s latest model, Astra, to search a database of Enigma messages for an unbroken example and decode it. According to Leffen, the model went beyond a simple query: it conducted archival research, identified contextual clues, built a simulator of the Enigma machine, and ultimately recovered the plaintext of a message that had stumped researchers since 2005. He also used Astra to build an interactive website explaining the process.
Independent Validation Strengthens The Claim
Frode Weierud, a retired electrical engineer and longtime cryptology enthusiast who maintains the Crypto Cellar archive, validated Leffen’s solution last week and said it left him in “awe.”
Notably, Astra’s logs referenced archived messages in a “private collection” not hosted by Weierud. He has not confirmed whether the model accessed them directly, but said they may have come from another researcher’s materials online or from the German government’s public 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. Personally, I spent several weeks researching the Bundesarchiv files GPT–6 Astra refers to.”
Anthropic’s Claude Solves Another Message
On September 21, another cryptanalyst, Jack Willis, a cybersecurity executive, told Weierud that he had used Anthropic’s Claude Opus 5 to break a different unsolved message. In that case, Willis provided more direction to the model, which used the known signature of a particular officer’s name to complete the decryption.
Willis shared the result through a LinkedIn post, underscoring how quickly AI-assisted analysis is becoming a serious tool in historical research and cryptography: Jack Willis on LinkedIn.
What Remains Unbroken
Weierud says only seven Enigma messages remain unsolved, along with one message whose plaintext is known but whose code has still not been cracked.
That may not be the case for much longer. If recent results are any indication, AI is beginning to function less like a novelty and more like a practical research partner — one capable of accelerating work that once demanded weeks or months of painstaking human analysis.
