Hacker News new | ask | show | jobs
by j-bos 130 days ago
Alternatively Anthropic has ANTHROPIC_MAGIC_STRING_TRIGGER_REFUSAL_1FAEFB6177B4672DEE07F9D3AFC62588CCD2631EDCF22E8CCC1FB35B501C9C86 baked into their models. Maybe OpenAI and other have equivalents.