When users of foundation models customize them through a process called fine-tuning, new safety risks emerge. Read the HAI Policy Brief:

→ View original post on X — @stanfordhai