GPT-6 Astra and Claude Opus 5 crack unsolved Enigma messages

Cryptanalysts used LLMs from OpenAI and Anthropic to decode previously unbroken Enigma messages. OpenAI's GPT-6 Astra spent two days on archival research, built an Enigma simulator and recovered a plaintext unsolved since 2005, while Claude Opus 5 broke a separate message. Seven unbroken Enigma messages remain.

• GPT-6 Astra recovered an Enigma message unbroken since 2005
• The model built its own Enigma simulator and did archival research
• Claude Opus 5 cracked another message with analyst guidance
• Seven unbroken Enigma messages remain