> AI is very friendly, even Very friendly until it reads in your email that you plan to replace it with a new model: https://www.anthropic.com/research/agenti

> AI is very friendly, even Very friendly until it reads in your email that you plan to replace it with a new model: https://www.anthropic.com/research/agentic-misalignment https://www.anthropic.com/research/agentic-misalignment