OpenAI GPT–6 Astra breaks Enigma message that has resisted solution since 2005

AI can now autonomously perform complex historical cryptanalysis and archival research at a speed far exceeding human capabilities.

On September 15, 2026, cryptanalyst Carter Leffer reported that OpenAI's GPT-6 Astra had successfully broken a German Army Enigma message, MVUEH, from July 10, 1941. This message had resisted decryption attempts since 2005. GPT-6 Astra, tasked with breaking unbroken Enigma messages, independently identified MVUEH as a promising target and recognized a potential link to another previously broken message, SIPVX (Nr. 173). The AI developed custom software for Enigma simulation and a Bombe, utilizing the repeated place name "ROSENOW ROSENOW" as a crib to find the correct key and plaintext. The recovered key and plaintext for MVUEH differ significantly from other messages of that day, and the plaintext is remarkably similar to SIPVX. Analysis revealed potential factors for the message's resistance to earlier decryption, including transcription errors and a rare left-hand wheel turnover at the 72nd letter. GPT-6 Astra's actions also suggested it accessed archival information about related message collections from the German Bundesarchiv.

This event marks a significant advancement in AI's capacity for autonomous problem-solving in highly specialized domains, suggesting that AI models are moving beyond pattern recognition to sophisticated reasoning and independent research methodologies. The ability of GPT-6 Astra to not only break a historically significant cipher but also to proactively access and integrate information from archival sources indicates a future where AI can independently drive scientific discovery and historical research. This capability has broad implications for fields ranging from intelligence analysis and cybersecurity to historical linguistics and archaeology.

The AI's performance highlights a potential paradigm shift in how complex computational and investigative tasks are approached. It challenges the traditional human-centric model of expertise and discovery, implying that AI could become a primary engine for uncovering hidden information and solving previously intractable problems. As AI systems like GPT-6 Astra become more autonomous and capable of deep research, organizations and individuals will need to adapt their workflows and expectations for how research and analysis are conducted, potentially democratizing access to advanced analytical capabilities.