AI Cracking Ciphers: Amateurs Solve 200-Year Historical Codes in Hours Using GPT-6 Astra

AI Cracking Ciphers Marks Turning Point for Historical Cryptography

AI is rapidly transforming historical cryptography, with large language models like GPT-6 Astra solving centuries-old ciphers that baffled experts for generations1

1

. Carter Leffen, a product development coach with no cryptography experience, used GPT-6 Astra to crack an 85-year-old Enigma message encrypted by the German army in July 19411

1

. The AI tested more than four billion settings in 10 hours, simulating the Enigma machine's complex encryption system that could scramble messages in up to 103 sextillion possible ways1

1

. Days later, Carter Church, a 27-year-old AI engineer at cybersecurity firm SentinelOne, deployed the same model to decode a Napoleon-era cipher from 1809 that revealed French troop positions2

2

. What would have taken weeks manually was solved in just six hours3

3

.

GPT-6 Astra and Claude AI Breaking Old Encryption Mysteries

The public release of state-of-the-art AI models from OpenAI and Anthropic has unleashed a wave of solutions to historical codes4

4

. Satoshi Tomokiyo, a Japanese patent engineer who maintains Cryptiana—an online list of unsolved ciphers—reports that since mid-September, most ciphers on his list have been cracked using AI3

3

. "2026 will be remembered as a turning point for codebreaking of historical ciphers," Tomokiyo wrote4

4

. Claude AI from Anthropic demonstrated similar prowess by tracing the origins of five NSA elliptic curve seeds in under 30 minutes—a mystery that persisted for over 25 years2

2

. The AI even questioned human cryptographers in its internal logs: "Is it suspicious that such an obvious thing wasn't discovered in 25 years?"2

2

.

How AI Tackles Historical Codes and Decryption Challenges

Large language models excel at code-breaking through persistence and computational power rather than traditional cryptanalysis expertise3

3

. "What AI mostly brings is persistence," Church explained5

5

. GPT-6 Astra employed simulated annealing—a common decryption strategy where keys are iteratively tweaked based on output comprehensibility1

1

. For the Enigma message, the AI wrote code to simulate the encryption machine and searched for settings that would render the message comprehensible, using a "crib"—a known word or phrase that might appear in the text1

1

. The models also leverage internet searches to find historical documents and partial keys that ease decryption1

1

. Church needed to learn "very little" about code-breaking techniques or the historical period to achieve his breakthrough4

4

.

Mixed Feelings as Technology Drains Satisfaction from Cryptography

While cryptography specialists welcome increased interest in historical codes, the AI surge brings complications1

1

. Elonka Dunin, a retired game developer and cryptography author, notes experts are overwhelmed by validation requests for AI solutions that are often incorrect: "There are so many false positives coming in from people who don't really understand cryptanalysis"1

1

. Tomokiyo expressed ambivalence about the rapid progress: "I'm excited long-time problems are now solved, but on the other hand, I have mixed feelings, because AI seems to have solved these difficult problems so easily"4

4

. One fellow code-breaker told him: "I feel useless"5

5

. Traditional methods remain effective but are time- and labor-intensive, leaving many ciphers unattended as historians tackle more pressing problems3

3

.

AI Transforming Historical Research Beyond Code-Breaking

The technology extends far beyond cracking ciphers, fundamentally transforming how historians access and analyze sources3

3

. AI is analyzing the Herculaneum Papyri—hundreds of scrolls carbonized in the 79 AD Vesuvius eruption—using X-rays to reveal hidden writing without physically unrolling the fragile documents4

4

. Italian researchers announced the first complete virtual unrolling of a scroll in June, uncovering a treatise on ethics5

5

. AI is building searchable databases by transcribing 10 million pages of Hebrew manuscripts for Jewish history research—"a huge Facebook of the past" according to project leader Daniel Stoekl Ben Ezra3

3

. At Notre Dame Cathedral archives, AI trained to transcribe 14,600 Latin pages achieves character recognition rates between 88 and 94 percent, revealing details about medieval manuscripts, daily cathedral life, building work, and care for abandoned children5

5

. Historian Julie Claustre notes that manual transcription would take "several decades," but automation "transforms the conditions of research" by enabling previously impossible inquiries5

5

. Some tough problems will likely remain unsolved due to insufficient relevant data for AI training3

3

.