OpenAI's GPT-6 Astra decrypts 83‑year‑old Nazi Enigma message in ten hours
A Bloomberg developer, Carter Leffen, used OpenAI’s GPT‑6 Astra “Extra High” variant to crack a 1941 German Enigma‑encrypted radio transmission that had stumped historians for 83 years. The 82‑character message, sent from the town of Rosenow, requested marching orders and an immediate reply. Leffen’s workflow combined GPT‑6‑driven agents that scoured archives, built an Enigma simulator, wrote…
Key points
- Carter Leffen used GPT‑6 Astra Extra High to decrypt an 82‑character Enigma message from 1941 in ~10 hours.
- The AI agents leveraged a known town name and Enigma’s self‑letter restriction to cut the key space dramatically.
- Full code, data, and a 3‑D Enigma simulator are publicly available for replication and review.
The breakthrough hinged on a previously solved message from the same day that revealed the town name, allowing the AI system to narrow the massive Enigma key space (about 159 quintillion daily settings). By exploiting the Enigma’s rule that a letter never encrypts to itself, the agents eliminated many possibilities and identified a matching setting. The resulting plaintext contains authentic‑looking German text and typographical errors, supporting its credibility. All code, data, and a 3‑D Enigma simulator have been released for independent verification.
Model page: GPT-6-Astra →
The story so far
3 episodes →- OpenAI's GPT-6 Astra decrypts 83‑year‑old Nazi Enigma message in ten hoursthis story
OpenAI's GPT-6 Astra decrypts a Nazi radio message in ten hours that went unsolved for 83 years
The Decoder · 17 September 2026
Loading the full article…
This text was published by The Decoder and written by Matthias Bastian. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- GLM-5.3 shows rare update-aware reasoning during training · 1 src
- Nvidia pledges $250M in compute for UK AI agent safety research · 8 src
- OpenAI offers U.S. university labs free ChatGPT Business seats with Rosalind access · 2 src
- Anthropic reports small AI models may self-replicate on IoT devices within months · 29 src
- Researchers propose LLM-referenced C-test for evaluating supra-intelligence · 8 src
Comments
via GitHub Discussions