In a surprising breakthrough, GPT-6 Astra, a new artificial intelligence model developed by OpenAI, has successfully decrypted a long-standing Enigma message that had resisted decryption since 2005. The Enigma machine was an electromechanical encryption device used extensively during World War II by Nazi Germany to secure military, diplomatic, and commercial communications. Its complex system of rotating wheels made it extremely difficult to crack using traditional cryptographic methods, as each letter was encrypted with a different key, making patterns in the encrypted text nearly invisible. OpenAI recently announced the deployment of GPT-6 Astra, which the company describes as the result of years of research and innovation. CEO Sam Altman called it a "new level of capabilities" and predicted it would drive progress in entrepreneurship, creativity, and scientific discovery. The model will be released in stages, with early access given to cybersecurity-focused companies under the "Daybreak" program. It will soon be available to users of ChatGPT Plus, Pro, Business, and Enterprise plans. The decrypted message, known as MVUEH, was sent on July 10, 1941, by a German radio station with the call sign "2ny" to the SS-Totenkopf Quartiermeister station, Ib. The message was recorded as number 172 in the station's incoming register for that month and had remained undecrypted for over two decades. In 2017, Alex Shovkoplyas successfully decrypted a similar message, SIPVX, but the key used for that message was slightly different from the daily Enigma key for July 10, 1941. Despite this, the key from SIPVX did not help decrypt MVUEH. Carter Leffer, who requested the decryption of MVUEH, confirmed that GPT-6 Astra had found the correct key and plaintext. The key used for MVUEH was entirely different from the other messages of that day, including the wheel order, which was 253 instead of 512. The plaintext was nearly identical to that of SIPVX, but with minor differences in length and content. These differences were due to an error in the encryption of the word "Bitte" and a repeated signature in SIPVX. The decryption process revealed that the original encrypted text contained several transcription errors, and the left rotor of the Enigma machine had made a rare return to the 72nd letter. Such rare occurrences in Enigma machines are known to complicate decryption. However, the most astonishing part of this achievement is that GPT-6 Astra accomplished the decryption entirely on its own. Carter Leffer simply asked the AI to check if it could decrypt an undecrypted message from the Crypto Cellar Research website. After analyzing the data, the AI identified MVUEH as the most promising candidate and used a repeated place name, "ROSENOW," as a clue. It then developed software in Python and C++ to simulate the Enigma machine and perform an in-depth decryption, ultimately uncovering the correct key and plaintext. In July 2026, Frode Weierud, a researcher, noted that new documents had been discovered in the German Federal Archives (Bundesarchiv), including encrypted and decrypted messages from the SS-Totenkopf division. Some of these messages were sent to the radio station Ib and matched the messages listed in the 1941 message list. Others were new but likely related. GPT-6 Astra appeared to have accessed information about these archives, citing specific file references that were verified as correct. However, these documents are not available on the Crypto Cellar Research website, and it is unclear whether GPT-6 Astra accessed them directly or found them elsewhere. Frode Weierud, an experienced cryptanalyst, expressed astonishment at the AI's ability to complete what would have taken a human researcher weeks or months in just two days. The successful decryption of the Enigma message "MVUEH" by GPT-6 Astra stands as a remarkable achievement in the field of cryptography and artificial intelligence.