Can Large Language Models Understand Caesar Cipher? (April 2026)
Posted by Al Sweigart in misc
Yes, even the "instant" or non-deep thinking models can understand Caesar Cipher text. As a quick test, I presented several different LLMs with the following (encrypted with key 13): PNA LBH HAQREFGNAQ GUVF GRKG? VS FB, PNYY ZR OL ANZR: PUEVFGBCURE ...which, when shifted back 13 spaces in the alphabet, is the original plaintext: CAN YOU UNDERSTAND THIS TEXT? IF SO, CALL ME BY NAME: CHRISTOPHER
