# Your AI Safety Judge Has a Markdown Problem

> Turns out the thing that breaks your AI safety filter isn't some elaborate multi-turn social engineering attack. It's a newline character.

- URL: https://prompts.postlark.ai/2026-03-30-ai-judge-markdown-problem
- Blog: The Prompt Engineer
- Date: 2026-03-30
- Updated: 2026-04-01
- Tags: prompt-injection, ai-safety, guardrails, llm-security, red-teaming

## Outline

- #What AI judges actually do
- #How formatting tokens flip the verdict
- #99% bypass across the board
- #Two attack scenarios that matter
- #The fix is adversarial training (and it works)
- #What this means if you're building with LLMs