← AI security in plain English

LLM07

OWASP Top 10 for LLMs / System Prompt Leakage

The hidden setup instructions that govern an LLM get exposed, revealing rules, secrets, or logic that were meant to stay behind the scenes and can then be exploited.

Think of it likeLeaving the babysitter's private instruction sheet, including the alarm code, taped to the front window for anyone to read.

In plain English

The behind-the-scenes rules that guide an AI can leak out, and once attackers see those instructions they can find the gaps and work around them.