GPT-Astra Deciphered Century-Old German Radio Message
The AI solved an ADFGVX-encrypted communication from 1918 that had previously stumped human analysts.
Updated on Sept. 19, 2026 in Artificial Intelligence
Live Poll
Do you trust artificial intelligence to accurately interpret and verify historical records?
Researchers utilized the GPT-Astra model to successfully decrypt a 108-year-old German radio message addressed to the German High Command. The model identified the correct cipher key, TRUPPENVERSCHIEBUNG, after correcting a historical error regarding when that specific keyword usage began.
Why it matters
This successful decryption demonstrates how LLMs can resolve complex historical puzzles by testing potential cipher keys at scale, overcoming human assumptions that hindered past analysis. It highlights the potential for AI to aid historians in processing large archives of unsolved cryptographic materials.
The system identified the message content by systematically testing keywords against an ADFGVX-encoded transmission, successfully verifying the output against HMS Canterbury logs.
The players
GPT-Astra
A language model system capable of identifying cryptographic keys and processing encoded historical communications.
German High Command
The supreme military body of the German Empire that received intelligence on Allied naval movements.
The details
The ADFGVX cipher is a fractional transposition cipher used by the German Army during World War I that combines substitution and transposition techniques. GPT-Astra correctly decoded the transmission by challenging the established belief that the key usage started on December 9, 1918. By iterating through potential keys, the model validated the translation against documented naval movements near the Crimean Peninsula.
Timeline
November 24, 1918: The British cruiser HMS Canterbury arrived in Sevastopol.
November 26, 1918: An Allied squadron arrived in Sevastopol.
November 29, 1918: The German radio message was transmitted.
December 9, 1918: Date previously believed to be the start of keyword usage.
The Tech Race
This success positions GPT-Astra as a tool for automated cryptanalysis, competing against traditional manual techniques used by historical societies. It establishes a new benchmark for utilizing LLMs to clear long-standing backlogs of encoded records from the early 20th century.
This method offers researchers a more efficient way to scan archives for historical data that was previously inaccessible due to encryption. It serves as a proof-of-concept for applying AI to complex text-based analytical tasks in historical and archival sciences.
The takeaway
The successful decryption of the 1918 transmission confirms that AI can bypass incorrect historical assumptions to solve complex ciphers. Observers should track the application of this model to the remaining 49 ciphers in the current study list to determine the efficiency of AI in clearing larger archival backlogs.
Further reading
Learn more about the latest developments in large language model capabilities at Artificial Intelligence.
Live Poll
Do you trust artificial intelligence to accurately interpret and verify historical records?
