AI Safety Guardrails Gone Wrong: Request Rejected for Being 'High Risk'
eptwts · x · 2026-08-11
Prominent researcher Andrej Karpathy shared a screenshot of what he calls the "best AI-generated reply" he has ever received.
The image shows the AI's safety mechanism being triggered, bluntly replying, "The request was rejected because it was considered high risk." This amusing failure of over-sensitive guardrails highlights the quirks of model alignment.
More from Fun
- Bun Creator Jarred Sumner Got Claude Results Using "Believe in Yourself" Prompts — ctjlewis · 2026-08-11
- Rotpilot: The CLI Tool That Blocks Brainrot Reels Until Claude Needs You — victorialslocum · 2026-08-11
- AI Recreates Monty Python's Pet Shop Sketch Using Video Generation — Boogertwilliams · 2026-08-11
- MiniMax Generates Cinematic Camera Transitions Using a Single Prompt — heypearlai · 2026-08-11
- AI-Generated Avatar Fan Film Perfectly Recreates Glowing Eyes Effect — eyishazyer · 2026-08-11
- Happy Horse 1.1 Model Generates Ultra-Realistic 1-Minute Pirate Sea Battle Short Film — azed_ai · 2026-08-11