AI decoding medieval love letters and mysterious messages

In a dark archive of an old Vatican library, a mysterious book lay untouched for centuries. Its pages were filled with strange symbols and unfamiliar signs that no one could understand. For more than 400 years, the manuscript seemed to conceal the stories of a lost world.

The book was known as the 'Borg Cipher'.

This 408-page manuscript contained 34 unusual symbols, some Roman letters, and an opening page written in Arabic. Over time, several pages were damaged, making its mystery even deeper.

Many believed the manuscript contained secret medical treatments for various diseases. During that era, discussing such subjects openly could be dangerous, as people could easily be accused of practicing witchcraft or sorcery.

For years, scholars tried unsuccessfully to decipher the text.

Eventually, artificial intelligence (AI) came to the rescue.

Researchers led by 'Beáta Megyesi' at 'Stockholm University' used machine-learning technology to crack the code. Gradually, they discovered that the manuscript contained descriptions of unusual medical remedies.

One passage suggested that drinking several glasses of high-quality red wine could help treat certain illnesses. Another recommended fermenting nutmeg in dough and consuming it as a remedy for dysentery.

For researchers, the process felt like a detective story. Every symbol, word, and pattern represented another clue toward unlocking a forgotten past.

Megyesi said the project revealed how much of human history remains unknown. Around one percent of documents preserved in libraries and archives worldwide are written partly or entirely in coded language.

These hidden texts contain stories about politics, wars, romance, medicine, secret societies, and personal relationships.

Sometimes, they can even reshape our understanding of history.

For example, researchers recently deciphered coded letters written by Mary, Queen of Scots. The letters revealed that while imprisoned, she was still plotting to reclaim her throne. They also suggested that her relationship with her son, James VI and I, was far from harmonious.

Not all codes are alike.

Some are simple, with one symbol representing one letter. Others are highly sophisticated, using multiple symbols for the same letter and even inserting fake symbols to confuse potential codebreakers.

A team led by French researcher 'Cécile Pierrot' spent six months deciphering a 500-year-old letter written by Charles V.

The letter contained 120 different symbols.

When researchers finally decoded it, they discovered that the powerful emperor feared an assassination plot against him.

One of the greatest challenges in this field is converting ancient handwriting into digital text. Many manuscripts are faded, damaged, or written in styles unfamiliar to modern readers.

Here too, AI is proving invaluable.

Researchers at the 'University of Oslo', including 'Michele Waldispühl', use an AI platform called Transkribus. The system can recognize historical handwriting and transform it into digital text.

The technology was used to analyze a secret 17th-century letter written during the devastating Thirty Years' War.

AI first separates the lines of text and then analyzes each character before producing a digital transcription. Although human experts are still needed to verify difficult sections, the technology has dramatically accelerated the process.

Researchers are now developing even more advanced systems.

Their goal is to create AI capable not only of digitizing texts but also of breaking codes directly from images.

Recently, they experimented with the 'Copiale Cipher', an 18th-century encrypted manuscript describing the rules and rituals of a secret German society.

Researchers showed the AI both the original handwriting and its translation. The system then learned to interpret sections it had never seen before.

This technology could unlock even greater mysteries in the future.

Scholars are currently collecting hundreds of old encrypted postcards, some of which contain hidden love letters. More than a century ago, people may have encoded their romantic messages to keep them secret from prying eyes.

Perhaps most remarkably, researchers are building an AI chatbot designed to help people decipher historical coded texts. Users could simply upload an image of an encrypted document, and the AI would attempt to read and translate it.

The system needed only 29 minutes to analyze part of the Borg Cipher. Beyond translation, it can also explain why a particular interpretation is likely correct—an important safeguard against AI-generated errors.

For researchers, the most exciting possibility lies elsewhere.

They believe AI may eventually decipher languages and scripts that no one can read today.

Ancient mysteries such as the Phaistos Disc, dating back around 4,000 years, and several undeciphered Greek inscriptions may one day yield their secrets.

If that happens, lost voices from the distant past could speak once again, and many of history’s closed doors may finally be opened.