Cryptanalysts say Astra and Claude Opus 5 decoded Enigma messages
Two cryptanalysts say they decoded World War II Enigma messages using OpenAI’s Astra and Anthropic’s Claude Opus 5. One of the solutions was validated by Frode Weirerud, who maintains an archive about the subject.
Why it matters · editorial interpretation
The case suggests that language models may accelerate historical and cryptanalytic research, but human validation remained necessary: Frode Weirerud confirmed one of the solutions. The origin of some references used by Astra remains unknown, an important issue when assessing the process’s reliability and reproducibility.
Two cryptanalysts say they decoded Enigma messages that had remained unsolved since World War II with help from large language models. Carter Leffen used OpenAI’s Astra to research archives, simulate the Enigma machine and recover the plaintext of a message that had puzzled researchers since 2005; Frode Weirerud, a retired engineer who runs the Crypto Cellar website, validated the solution. Jack Willis also said he decoded another message using Anthropic’s Claude Opus 5, guided by the known signature of an officer.
Weirerud says seven Enigma messages remain undecoded, along with another whose plaintext is known but whose code has not been broken. He said Astra achieved in two days what would take a human researcher weeks or months, although he still does not know how the model found references to files from a private collection.
Sources: TechCrunch