Jim Sanborn, creator of the Kryptos sculpture at CIA HQ, says he's frustrated with amateurs falsely claiming they cracked the 35-year unsolved code using AI
Jim Sanborn's inbox is flooded with amateur cryptographers who say they've cracked the code with chatbots like Grok 3.
Context & Ripple Effects
Kryptos remains an unresolved puzzle tied to CIA HQ, and its creator is now confronting a wave of chatbot-assisted claims that do not meet his standard for a solution.
The episode sits alongside coverage of the CIA using an AI chatbot to help analysts model foreign leaders, highlighting a distinction between AI as an analytical aid and AI output as proof of a factual claim.
First-order effects
- Sanborn must spend time rejecting and assessing a larger volume of purported solutions, while amateur claimants face a clearer message that a chatbot response does not establish that Kryptos has been solved.
- The public status of the puzzle remains unchanged: AI-generated assertions without a verifiable derivation are treated as false claims, not a breakthrough.
Second-order effects
- Cryptography communities and puzzle audiences have added incentive to demand reproducible methods, intermediate reasoning, and independent verification rather than screenshots or model-generated prose.
- Chatbot makers may face more visible examples of users mistaking fluent output for validated analysis, particularly in domains where the correct answer is unknown or difficult to independently check.
Third-order effects
- If this pattern persists, high-profile unsolved problems will increasingly become tests of AI epistemic reliability: the differentiator will be auditable evidence around a model's answer, not the answer's confidence or polish.
- The boundary between AI-assisted exploration and a validated discovery is likely to become more formalized in expert communities, though the pace will depend on whether users adopt shared verification norms.
The trend: Generative AI is broadening access to analytical tools while making verification, provenance, and reproducibility more important for claims of discovery.