ChatGPT-6 Astra Cracks 85-Year-Old German Enigma Message in Two Days

The MVUEH message was sent by the German Army to the SS-Totenkopf (Death’s Head) Division on July 10, 1941, at the height of Nazi military power in World War II. The encrypted cipher had been shared with enthusiasts online in 2005, but remained unsolved until Astra tackled it.

Crypto Cellar researchers report that Astra selected the message unprompted after being asked whether it could break any of the unbroken Enigma messages published on their web page. The AI picked Nr. 172, MVUEH, hypothesising it might relate to plaintext from another message, Nr. 173, SIPVX.

Working with a probable plaintext clue of the repeated place name "ROSENOW ROSENOW", Astra developed Python and C++ software for an Enigma simulator and an Enigma Bombe — the electromechanical device used by Allied codebreakers during the war. Its hunch proved correct.

The original 82-letter cipher, ICRVSORMCCWQTATYEVFXDBZGGSNXWLPSYWZYTCBSWULRTBZCVGODVJUSLSOOMJQJZSXSEBZPEYMDNXJYTC, decrypted to: "BTTE UM ANGABE DES MARSQWEGES X BEFINDE MIQ IN X ROSENOW ROSENOW X SOFORT FUNKANTWORT X WASCHBBSCH". An approximate English translation reads: "Please specify the route of march. I am in Rosenow, Rosenow. Immediate reply by radio."

Researchers are still analysing logs to determine how Astra achieved the feat so quickly. Notably, the AI performed the task "entirely on its own", according to the source, without human guidance beyond the initial question.

The breakthrough demonstrates a significant advance in AI’s ability to tackle complex historical cryptanalysis, though the precise methodology remains under investigation.

Analysis

Why This Matters

Demonstrates AI’s ability to autonomously solve complex, real-world cryptanalytic problems that have eluded human experts for decades.
If AI can replicate such feats reliably, it could accelerate research in cryptography, historical document recovery, and even modern codebreaking.
Raises questions about the transparency and interpretability of AI-driven problem-solving — researchers do not yet fully understand how Astra arrived at its solution.

Background

Enigma was the German cipher machine used extensively during World War II. Its encryption was considered unbreakable until Allied efforts, including work by Alan Turing and the Bletchley Park team, developed methods to decrypt messages. This particular 1941 message, MVUEH, had remained uncracked for 85 years despite being publicly shared among enthusiasts since 2005. ChatGPT-6 Astra is an advanced AI model capable of writing and executing code, which it used to simulate an Enigma machine and a Bombe-like attack.

Key Perspectives

Crypto Cellar researchers: The achievement is remarkable because Astra did it in two days with no human assistance beyond the initial request. They are still studying how the AI arrived at the solution, noting it would have taken a human weeks or months. Historians and cryptographers: The case study provides a valuable demonstration of AI’s potential in historical cryptanalysis, but caution that reliance on black-box AI methods could pose verification challenges for rigorous scholarship. Critics/Skeptics: Some may question whether the approach generalises to other unsolved historical ciphers, or whether the AI’s success relied on specific clues (such as the known plaintext guess). The lack of full explainability may limit trust in the method.

What to Watch

Publication of the full analysis logs from Crypto Cellar on how Astra built and executed its attack.
Attempts by other researchers to replicate the result using different AI models or manual methods.
Whether Astra or similar AIs can crack additional unsolved Enigma messages that remain publicly available.

Analysis

Why This Matters

Background

Key Perspectives

What to Watch

Sources