LLMs respond differently to harmful prompts when AI watermarking is used

2 points | by joozio 8 hours ago