A researcher used GPT-6 Astra to decipher a WWI German radio transmission from 1918, one of the 50 famous unsolved ciphers listed on a German science blog

AI models can now break ciphers thought to be too complex for automated solutions.

A researcher has utilized GPT-6 Astra to decipher a previously unsolved German radio transmission from World War I, dated November 27, 1918. The message, encoded using the ADFGVX cipher method, was one of 50 famous unsolved ciphers listed on the German science blog Scienceblogs.de. GPT-6 Astra identified "TRUPPENVERSCHIEBUNG" as the encryption key, a key that historical analysis suggests was not in use until December 9, 1918, explaining why the message eluded previous codebreakers. The decoded message reports the arrival of an English cruiser in Sevastopol and an allied squadron following shortly after. The model's findings were corroborated by historical logs, confirming the arrival of HMS Canterbury and an allied squadron around the dates mentioned in the transmission.

The successful decryption of a WWI cipher by GPT-6 Astra signifies a major leap in AI's capability for complex pattern recognition and historical data analysis. This accomplishment goes beyond typical language translation or content generation, demonstrating AI's potential to solve intricate cryptographic challenges previously requiring human expertise and significant computational resources. It suggests that AI models are evolving to tackle more abstract and rule-based problems, opening new avenues for historical research, intelligence analysis, and even cybersecurity.

This breakthrough has implications for both the study of history and the future of codebreaking. For historians, it offers a new tool to unlock previously inaccessible primary source materials, potentially rewriting understandings of past events. For the field of cryptography, it highlights the increasing need for post-quantum encryption methods and advanced AI-driven security measures to stay ahead of sophisticated decryption capabilities. The ability of AI to identify contextual discrepancies, like the date mismatch of the encryption key, also points to advanced reasoning capacities within these models.