How AI helped researchers decode a 217-year-old military letter linked to Napoleon Bonaparte, revealing details about French military planning in 1809.
By Adnan Mirza | October 13, 2026
AI Cracked Napoleon's 217-Year-Old Secret Code: What Did It Reveal?
Imagine finding a letter that has remained unread for more than two centuries. It contains hundreds of strange symbols, and the message inside could reveal important details about an army preparing for war.
For years, historians and code experts struggled to understand one such letter from the time of Napoleon Bonaparte. Then, in 2026, an artificial intelligence system helped solve the mystery in about six hours of work, according to the engineer behind the project.
The letter dates back to 1809. It contains military information linked to Napoleon's plans before a major campaign against Austria. The message describes army positions and gives instructions to a French commander who needed to understand what was happening across Europe.
But there is an important detail. Napoleon ordered the message to be sent, but he did not personally write the surviving coded letter. His stepson, Eugène de Beauharnais, sent it to Marshal Auguste de Marmont.
The Secret Letter That Survived for 217 Years
During the early 1800s, Europe was full of political tension and military conflict. Napoleon controlled a large part of the continent, but several countries stood against French power.
In 1809, war with Austria was approaching. Napoleon needed his commanders to know where French and allied troops were positioned. They also needed instructions about what to do if enemy forces began to move.
Sending these details in ordinary language was risky. If an enemy captured the messenger, the letter could reveal valuable information about French military plans.
To reduce this danger, Napoleon ordered the message to be written in code. The letter was sent from the headquarters of Eugène de Beauharnais, Napoleon's stepson and the Viceroy of Italy, to Marshal Auguste de Marmont.
Marmont was serving in Dalmatia, on the eastern side of the Adriatic Sea. He needed information about French and allied forces in other parts of Europe.
The message was important, but the code made it difficult for anyone without the correct key to understand it. Over time, the letter became a small but difficult puzzle for historians and cryptography experts.
Why Was the Code So Difficult to Break?
The letter was not written in a normal alphabet. It contained around 1,300 coded signs, with about 155 different signs used throughout the message.
These symbols formed part of a system known as a homophonic substitution cipher. In simple words, the same letter could be represented by more than one symbol.
For example, imagine that the letter E could be represented by three different signs. A person trying to break the code could not simply look for one symbol and replace it with E every time.
That made the work more difficult. The person solving the message had to identify repeated signs, understand how the code worked, and find the right way to turn the symbols into readable words.
There was another problem. The publicly available image of the letter was a low quality scan published in a French military journal in 1969.
The writing was not perfectly clear. Some symbols could be easy to confuse with others. Before the message could be decoded, the signs had to be read and grouped correctly.
For a human researcher, this could mean many hours of careful work. A single mistake in the transcription could cause problems later in the process.
How Artificial Intelligence Solved the Mystery
The breakthrough came when Carter Church, an AI engineer at cybersecurity company SentinelOne, decided to test whether an AI system could solve the historical cipher.
He used a model called GPT-6 Astra and gave it a scan of the coded letter along with the goal of solving the message.
Instead of relying only on guesses, the system had to work through several connected tasks.
Step 1: Reading the Symbols
The first challenge was to identify the signs in the old image. The AI examined the symbols and worked to distinguish different signs from repeated versions of the same sign.
This mattered because the cipher could not be solved reliably unless the original message was transcribed with care.
Step 2: Finding an Older Code Key
One of the most important steps was finding earlier research that had not previously been linked to this particular letter.
A cryptology historian named Daniel Tant had documented a partial table that explained 33 of the symbols. This covered roughly one third of the coded message.
The partial table did not explain every symbol, but it gave the AI a useful starting point. By connecting the table to the letter, the system could begin working out the meaning of the remaining signs.
Step 3: Reconstructing the Message
The AI used the available code information to work out the other symbols and translate the message into readable text. It also looked at historical information to check whether the result made sense.
According to Church, the work took about six hours. That included reading the image, researching the cipher, developing a solution, and checking the result against historical records.
The result was not accepted simply because an AI system produced a convincing story. Independent review helped check whether the decoded message and its key were consistent.
What Did Napoleon's Secret Message Say?
The decoded letter dealt with the positions and strength of French and allied military forces. It also explained the wider military situation as Austria prepared for war.
The message discussed troops in different regions and gave Marmont information that could help him understand the forces available to France and its allies.
One figure mentioned in the letter was an army of Friuli said to number 80,000 men. It is important to understand that this was a claim made in the historical message, not an independently confirmed count of soldiers actually ready to fight.
The letter also included advice about how Marmont should respond when he received orders to march. In effect, he was told not to be discouraged by a small enemy force or a group of poorly organised opponents.
This gives us a better idea of how Napoleon's military system worked. Commanders needed information from far away, and they had to make decisions based on reports sent across long distances.
The message was not a personal diary or a confession by Napoleon. It was a military communication containing information that France did not want its enemies to read.
A Clue Helped Confirm the Decoded Letter
One question naturally comes to mind. How could researchers know that the AI had found the correct message rather than creating words that only looked believable?
The answer lies in another historical document.
Napoleon's original correspondence included instructions dated March 16, 1809. In that communication, he asked Eugène to send Marmont a coded report about military positions and the wider situation in Europe.
This gave researchers a valuable way to check the newly decoded message. They could compare its contents with what was already known about Napoleon's instructions.
The comparison supported the reading of the letter. It also helped researchers place the document in the correct period.
The letter had previously been associated with 1807 in its published record. The new investigation linked it to March 1809, shortly before Austria began its offensive.
Independent review was also important. Satoshi Tomokiyo, who maintains the historical cipher website Cryptiana, checked the proposed solution and listed the cipher as solved.
This is a key part of the story. AI can offer an answer, but historians still need evidence, dates, documents, and other checks to judge whether the answer is reliable.
Did AI Discover a Secret That Changed History?
The result is impressive, but it is important not to exaggerate what it revealed.
The decoded message did not uncover a completely unknown war or change the accepted outcome of Napoleon's campaign. Much of its military information matched other historical correspondence that scholars already knew about.
Historian Michael Rowe of King's College London explained that the known correspondence helped researchers judge the translation. This made the result easier to check than a cipher whose contents were completely unknown.
The discovery still matters. For the first time, researchers could read the contents of this particular encrypted letter in full. It also helped restore its historical date and gave researchers a more complete view of the communication sent to Marmont.
The recovered code information may also help scholars study other documents that used the same or a related cipher system.
So the main achievement was not finding a secret fact that rewrites history. It was using modern tools to make an old, unreadable military message accessible again.
What Does This Mean for the Future of Historical Research?
Old archives contain letters, reports, maps, and records that researchers have not fully studied. Some are difficult to read because of faded ink, damaged paper, unusual writing, or coded symbols.
AI may help researchers deal with these problems more quickly. It can recognise patterns, compare signs, organise information, and search for connections that might otherwise take a long time to find.
Imagine a large collection of military letters from the 1800s. If several letters use related codes, solving one of them may provide clues for reading the others.
The same general approach could help with historical records written in difficult handwriting or with documents that contain repeated patterns. However, each case depends on the quality of the evidence and the type of problem being studied.
There are also limits. AI can make mistakes when reading old documents. It can misunderstand a symbol or produce an answer that sounds correct but is not supported by the evidence.
That is why researchers must check AI results against original documents and trusted historical records. A tool that helps solve a mystery is useful, but it still needs careful human review.
The Napoleon cipher shows what can happen when modern computing is combined with years of historical research. An old clue that once seemed unhelpful can become valuable when viewed in a new way.
Frequently Asked Questions
Did Napoleon personally write the coded letter?
No. Napoleon ordered the message to be sent. The surviving letter was sent by his stepson Eugène de Beauharnais to Marshal Auguste de Marmont.
How long did it take to solve the code?
According to AI engineer Carter Church, the main decoding work took about six hours using GPT-6 Astra.
Was the message completely unknown before 2026?
The coded letter had remained unread, but much of its expected content was known through other historical correspondence. The new work made the coded version readable and helped confirm its date.
Can AI decode every ancient secret message?
No. Some codes need missing keys, more documents, or specialist knowledge. Others may remain impossible to solve with the evidence available. AI can help, but it cannot guarantee a correct answer.
Final Thoughts
A message written during the Napoleonic Wars stayed hidden behind strange symbols for more than 200 years. In 2026, an AI assisted investigation helped turn those symbols into a readable account of military planning in Europe.
The real lesson is not that AI has replaced historians or that every mystery can now be solved in hours. It is that new tools can help people examine old evidence in new ways.
Somewhere in an archive, there may be another forgotten letter waiting to be understood. The next discovery could come from a historian, a code expert, or someone who simply asks the right question and knows how to check the answer.
What do you think? Can AI help us uncover more secrets from history, or should old documents always be solved by human experts first? Share your thoughts in the comments.
Sources and Further Reading
For readers who want to learn more, these reports explain the decoding work and its historical importance.
- Live Science: AI decodes a 217 year old secret letter ordered by Napoleon
- Euronews: AI decodes a Napoleonic letter in six hours
- Fondation Napoléon: A historian's review of the 1809 letter
Article by Adnan Mirza. Historical details have been checked against published reporting and historical commentary. The reported decoding time comes from the engineer behind the project.
