AI Researcher Cracks Napoleon's Secret Code After 217 Years
Translated & summarized from Calcalist by baba
The story in 5 lines · by baba
- AI researcher Carter Church cracked Napoleon's 217-year-old secret coded letter.
- The letter, sent in 1809, contained military details about French and Austrian troop movements.
- Church used OpenAI's GPT-6 Astra system and simulated annealing for decryption.
- The AI model took six hours to decipher the 1,300-character message.
- Experts warn AI's code-breaking ability could be used for malicious purposes.
An AI researcher from the Israeli-American cybersecurity firm SentinelOne has deciphered a coded letter sent by Napoleon Bonaparte to one of his senior generals, which had remained uncracked for 217 years. The letter, sent in 1809 by Eugène, Napoleon's stepson, to General Auguste de Marmont, contained details about French and Austrian troop movements prior to the war between France and the Habsburg Empire. Its contents were previously unknown because the code, comprising 1,300 characters representing letters and ancient words, was lost.
Carter Church, an AI researcher at SentinelOne, revealed this week that he used OpenAI's GPT-6 Astra system to crack and transcribe the message. According to the deciphered text, Napoleon wrote, "Austria will attack simultaneously on all fronts. It is constantly rushing towards its ruin." The advanced AI model took six hours to transcribe the characters and cross-reference them with previous decryption attempts that had already deciphered 33 characters.
Church employed a technique called simulated annealing, a trial-and-error process, comparing possible word combinations with existing writings by Marmont. While manual decryption was possible, it was an arduous and time-consuming task that no cryptanalyst had managed to complete until now. Church was inspired to attempt the decryption after reading about another AI model's success in cracking a German army coded letter from 1941. However, experts have cautioned that the advanced code-breaking capabilities of AI models could potentially be used to expose anonymous online profiles.
Read the original at CalcalistMentioned
The same event, reported separately by each outlet. Open a few to compare what different newsrooms emphasize — and what they leave out.