Astra and Opus just passed Turing’s other test

Alan Turing, a pioneering computer scientist, is most famous for the Turing test, which probes the ability to differentiate between artificial and human intelligence. Yet his most consequential wartime challenge was breaking the Enigma cipher employed by Nazi Germany in World War II.

Two modern cryptanalysts now report that they have solved two previously unbreakable Enigma messages with the help of large language models developed by OpenAI and Anthropic.

OpenAI’s Astra deciphers a 2005 Enigma puzzle

Developer Carter Leffen asked OpenAI’s model Astra to scan a collection of Enigma communications for an intact message and produce its plaintext.

The model succeeded, performing its own archival investigation, gathering contextual hints, constructing a functional Enigma simulator, and ultimately extracting the plaintext of a message that had puzzled researchers since 2005.

Leffen also leveraged Astra to develop an interactive website that walks visitors through the entire decryption effort.

Frode Weierud, a retired electrical engineer with a long‑standing fascination for cryptography, curates the Crypto Cellar website, which offers a range of resources, records, and a catalog of Enigma messages.

Weierud confirmed Leffen’s result last week, declaring that the outcome left him “in awe.”

Because AI systems pursue leads aggressively, Astra’s logs reference archived messages from a “private collection” that Weierud does not maintain.

Weierud remains uncertain whether Astra accessed those files, suggesting they might have been shared by another researcher online or drawn from the German government’s publicly available archives.

GPT–6 Astra is behaving like a very professional cryptanalyst and archive researcher. 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.

Anthropic’s Claude Opus 5 solves another Enigma message

On September 21, cybersecurity executive Jack Willis contacted Weierud to report that he had employed Anthropic’s model Claude Opus 5 to solve a different unsolved Enigma message.

Willis gave Claude extensive instructions, and the model eventually exploited the known signature of a specific officer’s name to decode the message.

Weierud observes that only seven unbroken Enigma messages remain, and one additional message has a known plaintext but its cipher has yet to be cracked.

The remaining mysteries may soon be solved.