Nestled near the CIA headquarters in Langley, Virginia, stands Kryptos, a cryptographic enigma sculpted by Jim Sanborn in 1990. This intricately designed artwork contains four encoded messages—three of which have been deciphered, while the final sequence has resisted decryption for over three decades. Despite relentless efforts by professional cryptographers and amateur codebreakers alike, the last unsolved section remains an enduring mystery.
But if you ask Sanborn, one thing is absolutely clear: No, you’re not going to crack Kryptos with a chatbot.
As reported by Wired, the artist, who has created works for institutions like MIT and NOAA, has been bombarded with messages from individuals who are utterly convinced they’ve cracked the elusive K4 section using AI tools. However, these aren’t experienced cryptanalysts or dedicated puzzle enthusiasts who have spent years poring over ciphers. Instead, they’re everyday people who have simply plugged the code into an AI chatbot and taken its response as gospel.
Speaking with Wired, Sanborn described a noticeable surge in submissions—a frustrating development for someone who, at 79, has been receiving potential solutions for over 35 years. The volume of messages became so overwhelming that he introduced a $50 fee just to review them. But beyond sheer quantity, it’s the attitude of these AI-assisted “codebreakers” that truly irks him.
“The tone of these emails is completely different,” Sanborn noted. “People who used AI to ‘solve’ Kryptos are absolutely convinced they cracked it between bites of their breakfast.”
The messages he receives are dripping with misplaced confidence. Some highlights include:
- “I’m just a vet…Cracked it in days with Grok 3.”
- “What took 35 years—and even the NSA with all its resources—could not do, I accomplished in 3 hours before my morning coffee.”
- “History’s rewritten…no errors, 100% cracked.”
Anyone who spends time on social media, especially on platforms like X, has likely encountered these kinds of people. The ones who confidently post, “Just Grok it” or smugly drop a chatbot-generated response as if it were the final word on any given topic.
Sanborn, however, is not impressed. For the record, no AI-generated submission has even come close to solving Kryptos. But beyond that, he questions why these individuals take such immense pride in outsourcing the challenge. There’s no deep analytical effort or breakthrough thinking involved—just an overreliance on a tool that confidently hallucinates answers.
This phenomenon isn’t unique to Kryptos. A study published in Computers in Human Behavior last year found that people tend to over-trust AI-generated advice, even allowing it to override their own judgment and conflicting information. Worse still, this reliance on AI can diminish human-to-human interactions—perhaps because those who lean on AI too much become insufferably self-satisfied.
For now, the last Kryptos puzzle remains unsolved, much to the delight of cryptography enthusiasts. And as far as Jim Sanborn is concerned, it’s going to take more than a chatbot to crack it.
