AI cracks centuries-old ciphers in hours

Artificial intelligence is helping non-specialists solve historical ciphers, including a coded 1809 message ordered by Napoleon Bonaparte that had gone unsolved for more than 200 years.

Carter Church, a 27-year-old AI engineer at US cybersecurity firm SentinelOne, used OpenAI's GPT-6 Astra model to decrypt the letter to one of Napoleon's generals in about six hours. The message described French and allied troop positions in lines of symbols, numbers and letters and had been sitting in an online academic archive.

Church said AI brings “persistence” to code-breaking and that he needed to learn little about either cryptography or the historical period. He consulted Japanese patent engineer Satoshi Tomokiyo, who keeps an online list of unsolved ciphers, to verify the result.

Tomokiyo wrote on his Cryptiana website that 2026 would be remembered as a turning point for historical code-breaking. He said most ciphers on his list had been cracked with AI since mid-September, though some harder problems may remain unresolved for years because less relevant data is available.

Researchers and historians are also using AI to read large collections of historical material. The technology is being used to analyse X-rayed Herculaneum Papyri, transcribe 10 million pages of Hebrew manuscripts and process 14,600 Latin pages from Notre Dame Cathedral's medieval archives.

Daniel Stoekl Ben Ezra, who leads the Hebrew manuscript project at the Ecole Pratique des Hautes Etudes in Paris, said no human could process that quantity of data. Historian Julie Claustre, who leads the Notre Dame archive project, said automation enabled research questions that were not previously possible.