These protocols focus specifically on mitigating catastrophic risks, and complement our work on other areas of AI safety, including studying societal impacts and developing techniques like Constitutional AI for alignment with human values.

→ View original post on X — @anthropicai,