AI models from OpenAI and Anthropic help crack two Enigma messages
The new decrypts shrink the list of unsolved Enigma messages and show large models can aid hands-on historical codebreaking.
Two cryptanalysts report that OpenAI’s GPT–6 Astra and Anthropic’s Claude Opus 5 have deciphered separate Enigma messages, leaving seven still unsolved. Carter Leffen directed Astra to find and decode a message in an archive; the model researched historical records, built an Enigma simulator and recovered plaintext from a message that had resisted solution since 2005. Frode Weirerud later verified Leffen’s result. In a separate effort, Jack Willis used Claude Opus 5 to solve another message, guiding it with a known signature associated with an officer; he contacted Weirerud about the result on September 21. Weirerud says one further message has known plaintext, but its Enigma code remains unbroken.
Why it matters
These decrypts matter mainly for historians and codebreakers: messages that had resisted human effort for years have now yielded with help from general-purpose AI models, reducing the tally of fully unsolved Enigma traffic and adding to the historical record of wartime communications. The work also shows that models like Astra and Claude Opus 5 can support practical cryptanalysis, from researching archives to simulating cipher machines and testing hypotheses about message structure and signatures.
Signal or noise?
Does this story matter, or is it hype? Decide before you see what everyone else thinks.
Sources
- TechCrunch