Anthropic literally implemented gaslighting in LLMs as safety measure lmao Yes, that’s what editing your prompt for you to change the output without letting you know is Gaslighting and sabotage

→ View original post on X — @theahmadosman