OpenAI's GPT-6 Astra and Anthropic's Claude Opus have successfully decrypted long-unsolved World War II Enigma messages, demonstrating advanced autonomous research capabilities. GPT-6 Astra cracked the 1941 MVUEH message in just two days by coding its own simulator. Meanwhile, Anthropic's Claude autonomously discovered a CRISPR-like enzyme system in bacterial DNA. These breakthroughs arrive amid rising geopolitical tensions, as the White House requests that OpenAI and Anthropic withhold new models from the UK's AI Security Institute until US security reviews are complete.
Decryption of the MVUEH Enigma message
- ▪OpenAI's GPT-6 Astra model successfully decrypted an unsolved 1941 German Army Enigma message, known as the MVUEH message, within two days.
- ▪OpenAI's GPT-6 Astra model decrypted the 1941 MVUEH Enigma message sent to the SS-Totenkopf Division by conducting archival research, building its own Enigma machine simulator, and finding context clues
History and validation of the MVUEH message
- ▪Retired electrical engineer Frode Weierud validated the decryption of the MVUEH Enigma message on his website, Crypto Cellar, in September 2026.
- ▪The MVUEH Enigma message, sent to the SS-Totenkopf Division on July 10, 1941, had remained unsolved by human cryptanalysts since it was shared online in 2005.
- ▪Frode Weierud stated that GPT-6 Astra's archival research of Bundesarchiv files in two days would take a human researcher weeks or months to complete.
Claude Opus decryption of Enigma message
- ▪Cybersecurity executive Jack Willis contacted retired electrical engineer Frode Weierud on September 21, 2026, to report that Claude Opus 5 had broken an unsolved Enigma message.
- ▪Anthropic's Claude Opus 5 model decrypted a separate unsolved Enigma message after receiving guided prompts from cybersecurity executive Jack Willis on September 21, 2026.
Discovery of CRISPR-Like enzyme system
- ▪Stanford University associate professor Stanley Qi praised Anthropic's Claude discovery of a CRISPR-like enzyme system, while microbiologist Kevin Blake expressed skepticism that the system would rival CRISPR technology
- ▪Anthropic announced that its Claude model autonomously discovered a new CRISPR-like enzyme system in bacterial DNA after searching a database of DNA sequences for 21 hours.
Government testing of new AI models
- ▪The White House Office of the National Cyber Director requested that OpenAI and Anthropic withhold their new AI models, including GPT-6 Astra and Mythos 5.1, from the UK's AI Security Institute until US government testing is complete
- ▪UK AI Security Institute director Henry de Zoete confirmed the institute was denied access to Anthropic's Mythos 5.1 model, available only to select United States organizations, but had tested OpenAI's GPT-6 Astra before its release.
Debate over global AI governance
- ▪The United States rejected global AI governance at the United Nations Security Council, emphasizing domestic encouragement of AI development over international regulatory frameworks.
- ▪At the United Nations General Assembly, UK Prime Minister Andy Burnham advocated for global AI standards, while US President Donald Trump rejected global governance of AI.
Debatable claims
- ▪AI-driven archival research is a reliable substitute for human historical scholarship
- ▪The autonomous capabilities of frontier AI models do more harm than good
- ▪National security concerns justify withholding US AI models from foreign testing institutes
Story comments
Loading comments…