OpenAI's GPT-6 Astra successfully decoded a 217-year-old encrypted military letter from Napoleon in just six hours. The breakthrough reveals Napoleon's battlefield message to General Marmont detailing troop positions across Europe before Austria's 1809 invasion. Historians call this AI-driven breakthrough a modern Rosetta Stone for historical codebreaking.

AI Decodes 217-Year-Old Secret Letter From Napoleon's Era

Carter Church, an AI engineer at cybersecurity firm Sentinel One, has successfully used OpenAI's GPT-6 Astra to crack a previously unsolved cipher from Napoleon Bonaparte's military correspondence. The AI decodes letter sent in 1809 by Napoleon's stepson Eugène to General Auguste de Marmont, revealing military secrets that remained hidden for 217 years.

1

The breakthrough required just six hours and a single prompt, marking a significant milestone in historical codebreaking.

Source: Live Science

Source: Live Science

The secret letter was dispatched on Napoleon's orders in March 1809, less than two weeks before Austria invaded Bavaria to start a new war. Marmont commanded roughly 13,000 troops stationed in Dalmatia on the eastern shore of the Adriatic Sea, separated from other French forces by Austrian territory.

2

The message's purpose was to inform Marmont of every French and allied army position across the continent without revealing military secrets if intercepted.

How GPT-6 Astra Cracked the Homophonic Cipher

Church found the letter while searching for an appropriate code to demonstrate GPT-6 Astra's capabilities. The only publicly available version was a low-resolution scan published in a French army journal in 1969. The homophonic cipher contained 155 different signs totaling 1,300 symbols, where common letters had several interchangeable symbols to increase encryption complexity.

1

The AI model divided the document into rows, distinguished between handwritten symbols, and transcribed roughly 1,300 cipher units before reconstructing their meaning. During its web research, Astra independently discovered a partial table by cryptography historian Daniel Tant that documented 33 of the symbols, accounting for about a third of the letter. Tant had catalogued the table as belonging to correspondence between Eugène de Beauharnais and "Général Grammont," possibly a naming mistake for Marmont. GPT-6 Astra identified this connection and successfully applied it to decode the remaining symbols.

1

Napoleon's Secret Battlefield Message Reveals Military Strategy

The decoded message told Marmont, "I hasten today to acquaint you with the disposition of French and allied armies." It detailed that 40,000 Bavarians occupied Munich and Passau, 30,000 Polish troops threatened Krakow, and tens of thousands of French troops were stationed around Bayreuth, Ulm, Donauwörth, and Augsburg.

2

The letter also claimed "The Russians are marching against Austria" and warned that Austria was moving "towards its ruin."

The message encouraged Marmont not to fear opposition, stating: "When you receive the order to march, you must not allow yourself to be intimidated by a few troops or a gathering of rabble." It noted that Austrian troops "appear to threaten the Tyrol" but cautioned "There is, however, still nothing definite," while questioning how much militia reinforcements would strengthen the Austrian army. The message closed with "I continue to await news from you with impatience."

2

Historical Verification and the Rosetta Stone Comparison

The decoded content closely matches Napoleon's original order sent to Eugène on March 16, 1809, instructing him to send Marmont a ciphered letter explaining French and allied army positions. The newly decoded document lists the same forces in the same order with identical troop numbers, providing verification of the translation's accuracy.

2

Michael Rowe, a researcher in European history at King's College London, called the findings "impressive" and described the methodology as "a bit like a kind of Rosetta Stone." He explained that testing AI on this letter, where historians already knew the expected contents from Napoleon's original correspondence, builds confidence for using the same tools on documents where contents are unknown.

1

Zack White, a history research fellow at the University of Portsmouth and host of the Napoleonic Wars Podcast, noted the translation raises "some curious questions about Napoleon's mindset in the run up to the campaign in Austria in 1809." Napoleon actually had 60,000 fewer troops available than claimed in the letter, and appeared fixated on the number of troops Austria might send to invade Italy.

1

Wave of AI-Driven Breakthrough in Historical Codebreaking

George Lasry, a researcher specializing in historical cryptography, revealed that large language models solved more than 40 historical ciphers in September alone, working in an autonomous setup developed by data scientist Dan Bourdeau. In 24 cases, models reconstructed keys using only ciphertext. AI systems decoded encrypted letters sent in 1585 and 1586 by Henry of Navarre, the future King Henry IV of France, to an envoy raising a German army.

2

Another project decoded a 1916 telegram from Chinese revolutionary Sun Yat-sen's circle, expanding to roughly 60 intercepted messages including one reporting the assassination of revolutionary Chen Qimei. These systems perform historical detective work by searching enormous archives, identifying related documents, and testing clues before cryptanalysis. Lasry and Bourdeau are studying whether the models' advantage comes from stronger reasoning or their ability to search, test hypotheses, and repeat work at speeds no human cryptanalyst could match. Thousands of encrypted historical documents sitting unread in archives may soon be accessible.

2

Today's Top Stories

© 2026 TheOutpost.AI All rights reserved