Degrading Safety Alignment via OpenAI's Fine-Tuning API substack.com 1 points by sleepingreset 7 days ago