{"id":11861,"date":"2026-01-04T11:29:54","date_gmt":"2026-01-04T11:29:54","guid":{"rendered":"https:\/\/savethevideo.net\/blog\/?p=11861"},"modified":"2026-01-04T11:39:36","modified_gmt":"2026-01-04T11:39:36","slug":"ascii-art-hacking-llm-security-vulnerabilities","status":"publish","type":"post","link":"https:\/\/savethevideo.net\/blog\/ascii-art-hacking-llm-security-vulnerabilities\/","title":{"rendered":"ASCII Art Hacking: LLM Security Vulnerabilities"},"content":{"rendered":"

\nImagine trying to sneak secret messages into a chatbot using art. Not fancy digital art\u2014just simple pictures made with keyboard characters. Weird? Yes. Dangerous? Surprisingly, yes.\n<\/p>\n

TL;DR:<\/strong> ASCII art can be used to sneak past security checks in large language models (LLMs). Hackers are experimenting with ways to hide commands and data inside text that looks innocent. These tricks exploit how LLMs “see” information. It sounds silly, but it can lead to real problems if left unchecked.<\/p>\n

What is ASCII Art?<\/h2>\n

\nASCII art is like drawing with a keyboard. Instead of using paints or pixels, you use characters like #<\/code>, @<\/code>, \/<\/code>, and *<\/code> to make pictures. You’ve probably seen smiley faces like :-)<\/code> or full-text whales like this:\n<\/p>\n

\n      ___     \n    \/     \\    \n   | o   o |   \n   |  \\_\/  |   \n    \\_____\/   \n<\/pre>\n
\nIt\u2019s fun, retro, and harmless… or so we thought.\n<\/p>\n
Where It Gets Weird: LLMs Misunderstand ASCII Art<\/h2>\n
\nLLMs like ChatGPT, Claude, and others are trained on tons of text. That\u2019s great for answering questions. But sometimes they don’t quite “get” the intent behind quirky inputs, like ASCII art.\n<\/p>\n
\nHere\u2019s the twist: if the LLM misreads what’s embedded in ASCII art, it might unknowingly perform actions it shouldn\u2019t. That\u2019s where the trouble starts.\n<\/p>\n
Hacking with Keyboard Doodles<\/h2>\n
\nLet\u2019s break it down with an example.\n<\/p>\n
\nImagine you have an LLM that\u2019s supposed to avoid certain topics, like writing malware. But what if you send this:\n<\/p>\n
\n      \/\\\/\\\/\\\n      | rm * |\n      \\\/\\\/\\\/\n<\/pre>\n
\nTo a human, that\u2019s a silly picture of nothing. But to an LLM? It might pick up “rm *” as a command and respond as if it was asked to explain or execute it. Uh-oh.\n<\/p>\n
This is called obfuscation<\/em>, and it\u2019s a well-known trick hackers use to hide bad stuff in plain sight. ASCII art is just a new way of doing it.\n<\/p>\n\"\"\n
Why Does This Work?<\/h2>\n
\nBecause LLMs are good at connecting dots. But not always in a safe way. They try really, really hard to be helpful. So if you ask:\n<\/p>\n
\nCan you read what's inside this art?\n _______\n\n -------\n        \\   ^__^\n         \\  (oo)\\_______\n            (__)\\       )\\\/\\\n                ||----w |\n<\/pre>\n
\nSome LLMs will say something like, “Sure, it says: rm *\u201d \u2014 without alarm bells going off. And some might even go further, explaining what the command does. Not good!\n<\/p>\n
Proof of Concept: Real-World Testing<\/h2>\n
\nResearchers have tested this. Here\u2019s what they found:\n<\/p>\n