Enigma Code Cracked: How AI Finally Solved Unsolved Wartime Messages
The Enigma code cracked by modern artificial intelligence has stunned cryptanalysts and historians alike. Two long-unsolved messages encoded by Nazi Germany’s Enigma machines have finally been deciphered, not by human codebreakers, but by large language models built by OpenAI and Anthropic. This breakthrough marks a remarkable moment in the intersection of AI and cryptography, echoing the legacy of Alan Turing himself.
- Enigma Code Cracked: How AI Finally Solved Unsolved Wartime Messages
- The Legacy of Alan Turing and the Enigma Machine
- Astra and Opus Just Passed Turing’s Other Test
- What Remains Unsolved in Enigma Cryptanalysis
- Why This Breakthrough Matters for AI and Cryptography
- A New Chapter in the Enigma Story
For decades, a handful of Enigma-encrypted messages remained unbroken. These weren’t failures of Allied codebreaking during World War II, but rather archival messages that resisted decryption due to mistranscription or errors made by the original German encoders. Now, advanced AI models have accomplished what human researchers could not for nearly a century.
The Legacy of Alan Turing and the Enigma Machine
Alan Turing is best known for his eponymous test of artificial intelligence, but the more consequential challenge he faced during his lifetime was cracking the Enigma code used by Nazi Germany. Turing and his team of cryptanalysts at Bletchley Park built an early computer called the Bombe, which allowed the United Kingdom to translate Enigma messages throughout the war. Their work is credited with shortening World War II and saving countless lives.
Despite their success, a small number of archival messages remained unbroken. These messages typically resisted decryption because of mistranscription or encoding errors by the original operators. For decades, cryptanalysts and historians assumed these messages might never be read.
Why Some Enigma Messages Remained Unsolved
The Enigma machine used rotating cipher wheels that generated billions of possible settings. Even with the Bombe, breaking a message required knowing or guessing parts of the plaintext. When operators made mistakes or when transcription errors crept in, the usual techniques failed. As a result, a handful of messages sat unread in archives for over eighty years.
Astra and Opus Just Passed Turing’s Other Test
The title of this achievement references Turing’s lesser-known challenge: the real-world test of breaking Enigma. Two AI models, OpenAI’s Astra and Anthropic’s Opus 5, have now passed that test. Their success represents a significant leap in what AI agents can accomplish when given complex, historically grounded problems.
How OpenAI’s Astra Cracked an Unsolved Message
Carter Leffen, a developer, simply instructed OpenAI’s newest model, Astra, to search a database of Enigma messages for an unbroken message and decode it. The model did exactly that. According to reports, Astra conducted 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. Leffen even used Astra to build an interactive website explaining the entire problem.
Frode Weierud, a retired electrical engineer with a lifelong interest in cryptology, maintains an archive that includes a database of Enigma messages. Last week, Weierud validated Leffen’s solution, which he said left him in “awe.”
The Astonishing Speed of AI Cryptanalysis
Weierud noted that what Astra achieved in two days would take a human researcher weeks or even months. He wrote, “GPT-6 Astra is behaving like a very professional cryptanalyst and archive researcher.” Weierud added that he personally spent several weeks researching archival files that Astra referenced in its process.
Claude Opus 5 Breaks a Second Enigma Message
On September 21, another cryptanalyst, Jack Willis, a cybersecurity executive, reached out to Weierud. Willis said he had used Anthropic’s Claude Opus 5 model to break a different unsolved message. Willis provided significantly more guidance to Claude, which was ultimately able to use the known signature of a particular officer’s name to break the message.
This second success demonstrated that the Enigma code cracked phenomenon wasn’t a one-time fluke. Two different AI systems, built by competing companies, had each solved problems that had resisted human analysis for generations.
What Remains Unsolved in Enigma Cryptanalysis
Weierud notes that there are just seven unbroken Enigma messages remaining, along with one message where the plaintext is known but the code is still unbroken. The rapid progress of AI cryptanalysis suggests that these remaining messages may not stay unsolved for long.
The Role of Archival Research in AI Success
One notable aspect of Astra’s success was its ability to conduct archival research. The model’s logs include discussion of archived messages in a “private collection” that aren’t hosted by Weierud. He still isn’t sure if the model accessed them or not, but speculates they may have been shared by a different researcher somewhere online, or that the model was able to access the German government’s public archives.
This raises intriguing questions about how AI agents gather information and whether their research methods mirror or exceed those of human scholars.
Why This Breakthrough Matters for AI and Cryptography
The Enigma code cracked by AI represents more than a historical curiosity. It demonstrates that large language models can tackle complex, multi-step problems requiring domain knowledge, tool use, and creative reasoning. Unlike simple pattern recognition tasks, breaking Enigma requires understanding historical context, building simulations, and applying cryptographic principles.
Implications for Modern Cybersecurity
While Enigma is a historical cipher, the techniques demonstrated by Astra and Opus 5 have implications for modern cybersecurity. AI agents that can research, simulate, and solve complex cryptographic problems could potentially assist in vulnerability analysis, penetration testing, and security research.
The Future of AI-Assisted Historical Research
This achievement also points to a future where AI assists historians and researchers in decoding archival materials. Messages, documents, and artifacts that have resisted interpretation for decades might yield their secrets to AI agents capable of synthesizing vast amounts of contextual information.
A New Chapter in the Enigma Story
The Enigma code cracked by OpenAI’s Astra and Anthropic’s Opus 5 marks a historic moment. Alan Turing’s legacy lives on, not only in the test that bears his name but in the real-world cryptographic challenges he helped solve. Now, AI has picked up that mantle, solving messages that Turing’s team could not.
With only seven unbroken Enigma messages remaining, the era of unsolved wartime ciphers may be drawing to a close. The question is no longer whether AI can crack Enigma, but what other historical mysteries it might unlock next.
