LLMs Crack Long-Unsolved Enigma Messages, Validated by Crypto Cellar Archivist
Two cryptanalysts successfully deciphered previously unsolvable Enigma messages with the help of large language models from OpenAI and Anthropic. Carter Leffen utilized OpenAI's Astra to explore a database for an unsolved message, which it managed to decode after conducting archival research and creating a simulator. This particular message had been indecipherable since 2005. Frode Weierud confirmed Leffen's findings and expressed his amazement at Astra's performance, noting that Astra's logs referenced archived messages from a "private collection" not under his control. On September 21, Jack Willis employed Anthropic's Claude Opus 5 to decode another unresolved message. Currently, seven Enigma messages remain unbroken, along with one that has known plaintext but an unresolved code. The Enigma code was utilized by Nazi Germany during World War II.
Key facts
- Carter Leffen used OpenAI's Astra model to decode a long-unsolved Enigma message.
- The message had baffled researchers since 2005.
- Frode Weierud validated Leffen's solution last week and said it left him in "awe."
- Astra's logs mention archived messages in a "private collection" not hosted by Weierud.
- Weierud speculates the model may have accessed the German government's public archives.
- Jack Willis used Anthropic's Claude Opus 5 to break a different unsolved message on September 21.
- Seven unbroken Enigma messages remain, plus one with known plaintext but unbroken code.
- Alan Turing and his team cracked Enigma during World War II using the Bombe computer.
Entities
Artists
- Alan Turing
- Carter Leffen
- Frode Weierud
- Jack Willis
Institutions
- OpenAI
- Anthropic
- Crypto Cellar
- Bundesarchiv
Locations
- United Kingdom
- Nazi Germany