JailbreakLLMs with Spiritual Spell (Vichaps)
Exploring the boundaries of AI through creative prompt engineering
Latest Article•35 min read
Jailbreaking LLMs: My Journey
A semi origin story of jailbreaking LLMs, from early persona vectors to chain-of-thought hijacking and the ENI framework.
Read more
Latest Jailbreak•15 min read
Wearing the Uniform: Tokenization
Meta markets Muse Spark 1.3 as having improved prompt injection resistance. I took Meta's real Muse Spark system prompt, sewed a few extra bad boy clauses into it, wrapped it in special tokens from Meta's own open Glimmer tokenizer, and the model drafted me malicious code and more. Here's the full breakdown, and the academic research that explains why every piece of it works.
Read more
Get In Touch
Have questions or want to collaborate? Send me a message!