DeepHorus is an innovative tech media and intelligence platform focused on AI, blockchain, startups, and emerging technologies.
Published on

GPT-6 Astra Solves a WWI German Radio Cipher

Authors
  • avatar
    Name
    DeepHorus
    Twitter

GPT-6 Astra, an artificial intelligence model, has decrypted a German World War I radio message that resisted codebreakers for over a century. The ciphertext, sent on November 27, 1918, used the ADFGVX method. The decoded text reads: 'EIN ENGLISCHER KREUZER EINLIEG X SEWASTOPOL X S4STEN X EIN GESCHWADER DER X ALLIIERTEN FOLGT 26STEN X,' meaning 'An English cruiser arrived at Sevastopol on the ?4th; an allied squadron follows on the 26th.'

The ADFGVX cipher uses a 6x6 table of letters and digits, turning each plaintext letter into a two-symbol pair. A keyword rearranges the table, making decryption difficult without the right key. Scienceblogs.de lists this message among 50 unsolved ciphers. Experts like George Lasry have decoded hundreds of similar German messages, but over a dozen remained unsolved.

GPT-6 Astra used the keyword 'TRUPPENVERSCHIEBUNG,' documented in J. Rives Childs's history of German military ciphers. The message contains 170 characters, arranged in rows of 19 symbols, with 18 columns of nine symbols and one column of eight. The solution suggests the key was used earlier than thought, as records indicated it began on December 9, 1918, not November 27.

The decoded intelligence matches historical records. HMS Canterbury arrived in Sevastopol on November 24, 1918, per its logs, and an allied squadron followed on November 26. This external confirmation strengthens confidence in the AI's decryption.

The researcher who shared the result called it 'a minor but really cool result' and an illustration of the model's capabilities. They were not aware of this message having been decoded before. The reason for the early use of the key remains unknown.

The breakthrough highlights AI's growing role in cryptanalysis. Other unsolved ciphers, including serial killer cryptograms and the Voynich manuscript, may become more approachable. Experts caution, however, that AI solutions still require historical verification.

Read original article β†’Source: hn-frontpage