GPT-6 codebreaking AI cracks 217-year-old Napoleon cipher in six hours

What happened

A coded letter written for a general of Napoleon Bonaparte stayed unread for 217 years. Carter Church, a staff AI engineer at cybersecurity company SentinelOne, tested OpenAI's GPT-6 Astra on it by feeding the model one scanned image of the cipher and a prompt.

Confirmed

Global impact / market context

If AI models can break historical ciphers from a single image, the same pattern-reading skills could speed up code and data protection work, cutting the time and cost companies spend on security testing. That may shift spending toward AI security tools and away from slower manual methods.

Analyst inference

Security budgets and AI tool spending are linked: faster codebreaking by defenders can raise demand for AI security products and pressure older security vendors. For investors, companies tied to AI security may gain, while those selling slower manual testing face competition from cheaper AI alternatives.

Analyst inference

What to watch

1. The article says Carter Church is a staff AI engineer at cybersecurity company SentinelOne, so future reports may show whether SentinelOne turns this test into a product or service. Confirmed
2. Watch whether OpenAI or other AI developers publish rules for using GPT-6 Astra on ciphers, which means limits on codebreaking tasks, because new rules could change how security firms sell such tools. Proposed
3. Watch whether companies adopt AI codebreaking for their own security testing, since faster testing could lower security costs and shift spending toward AI vendors and away from older tools. Analyst inference

Affected assets

• AI — Any Inu

Evidence

• en.cryptonomist.ch