AI-assisted analysis has reportedly helped decipher a 1809 military letter sent by Eugène de Beauharnais to Marshal Auguste de Marmont. Napoleon ordered the information to be sent in cipher, but he did not write the surviving letter. The result offers a useful case study in how AI can aid historical cryptanalysis—and why a plausible, checkable solution is not the same as proof that AI can crack any unknown code.
Who sent the letter, and why?
In March 1809, as war with Austria approached, Napoleon instructed Eugène, then Viceroy of Italy, to tell Marmont where French and allied forces were positioned. Napoleon’s 16 March instruction specified that the information should be sent in cipher by a reliable officer. Marmont’s troops were in Dalmatia, separated from other French forces.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
Napoleon: A Life | $20.49 | Buy on Amazon |
| 2 |
|
The Campaigns of Napoleon | $59.93 | Buy on Amazon |
| 3 |
|
Napoleon: A Concise Biography | $17.48 | Buy on Amazon |
| 4 |
|
Napoleon Unleashed: A History of the Revolutionary, Emperor, and Military Genius who Reshaped Europe... | $17.99 | Buy on Amazon |
| 5 |
|
Who Was Napoleon? | $6.51 | Buy on Amazon |
As an Amazon Associate I earn from qualifying purchases.
The surviving ciphered message was sent from Eugène’s headquarters to Marmont. Its purpose was practical: report army positions and encourage Marmont not to be intimidated when he received orders to march. Calling it a letter “ordered by Napoleon” describes the instruction behind it, not its authorship.
What did the deciphered message say?
Church’s reported reading gives the positions of French and allied forces and urges Marmont not to be deterred by a small force or “a gathering of rabble” when ordered to march. Euronews reports that the text gives the army of Friuli’s strength as 80,000. That is a figure in the deciphered letter, not an independently verified count of troops actually present.
#1 Best Overall
The content broadly fits Napoleon’s separate correspondence, which already told Eugène what the report should cover. Michael Rowe, a researcher in European history at King’s College London, said that consistency makes the translation fairly credible, while noting that expected content provides a different test from deciphering a message whose subject is wholly unknown. Rowe also said the cipher repeats a large segment of Napoleon’s original letter, so it does not disclose a major new military secret. Live Science reported Rowe’s assessment.
How the AI-assisted decipherment reportedly worked
Carter Church describes the message as a homophonic cipher: a system in which a common plaintext letter can be represented by more than one sign. The surviving image was a low-resolution plate printed in a 1969 French army journal. According to Church’s account, the cryptogram contains 1,300 units represented by 155 distinct signs.
Rank #2
The solution was not built from a blank slate. Daniel Tant had already published a partial table that supplied 33 symbol-to-letter values, covering 435 units—about one-third of the ciphertext. Church reports working out the remaining 865 units, about two-thirds, with an AI-assisted workflow that transcribed the image, grouped variant handwritten marks, analyzed the cipher, wrote code, and checked proposed values against the image. He reported approximately six hours of model execution time; that is his account of this case, not an independent benchmark of AI cryptanalysis.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Church published a research record with the transcription, tables, and scripts. Euronews reports that the package was published on 18 September 2026; Church’s write-up lists a solution package versioned 1.0.0 and dated 19 September. The materials make the proposed solution checkable in principle, but publication of scripts is not the same as an independent reproduction of every reading. Church’s solution record describes the method and provides the materials; Euronews’ report gives its publication date.
Rank #3
Why experts call it a useful test—not a universal breakthrough
Rowe called the cipher “a bit like a kind of Rosetta Stone.” In this case, the partially known key and Napoleon’s related correspondence offer footholds for interpreting the remaining signs and checking whether the result makes sense. Church told Live Science that “The verifiability of solutions is actually the most difficult part, far more difficult than generating potential solutions.”
That distinction matters. A system may generate candidate readings, but historians still need to establish that the transcription matches the handwritten signs, that symbol assignments behave consistently across the text, and that the resulting message fits the documentary context without relying on that context alone. Here, Church reports image checks and a released key, while the known subject matter provides a strong contextual comparison. The available coverage does not establish a formal, independent audit of the full recovered key.
The case therefore demonstrates a reported use of AI-assisted tools on a particular cipher with partial prior work and contextual clues. It does not show that AI can reliably decipher any unknown cipher or unread historical script. As Zack White, history research fellow at the University of Portsmouth, told Live Science: “It’s important that historians now set the AI’s works in a bigger picture of the history of the period.” Live Science’s coverage includes the historians’ reactions and qualifications.
Quick Recap
Best Value
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
