OpenAI and Anthropic Models Crack Unsolved World War II Enigma Ciphers

Frontier artificial intelligence models built by OpenAI and Anthropic have decrypted two historical World War II Enigma machine messages that remained unbroken for decades, according to a report from TechCrunch AI (https://techcrunch.com/2026/09/25/astra-and-opus-just-passed-turings-other-test/). The results illustrate how large language models are executing complex cryptanalysis, autonomously researching archives, and writing specialized software simulators.

The first cipher was decoded by developer Carter Leffen, who prompted OpenAI’s Astra model to search an online database of historical Enigma dispatches, select an unsolved transmission, and break it. The model conducted archival research, identified context clues, constructed an Enigma machine simulator, and recovered the plaintext of a message that had defied researchers since 2005. Leffen subsequently used Astra to generate an interactive website documenting the solution.

Frode Weierud, a retired electrical engineer who curates the cryptology archive Crypto Cellar, validated Leffen’s result. Astra's execution logs referenced archival materials from a private collection not hosted on Crypto Cellar, leading Weierud to speculate that the model found digitized records on the open web or queried Germany's federal archives directly. "GPT–6 Astra is behaving like a very professional cryptanalyst and archive researcher," Weierud said. "What it has achieved in two days would take a human researcher weeks or even months. Personally, I spent several weeks researching the Bundesarchiv files GPT–6 Astra refers to."

In a separate breakthrough on September 21, cryptanalyst Jack Willis informed Weierud that he had used Anthropic’s Claude Opus 5 to crack another unsolved Enigma dispatch. While Leffen relied on an autonomous workflow, Willis provided Claude Opus 5 with directed guidance, utilizing the known signature pattern of a specific officer to isolate the cipher's settings.

During World War II, British cryptanalyst Alan Turing and his team designed the electromechanical Bombe to break Enigma transmissions, but a small collection of messages remained unsolved due to encoder errors and mistranscriptions. According to Weierud, only seven unbroken Enigma messages remain, along with one dispatch where the plaintext is known but the encryption key settings remain unsolved.