Sam had a blog post literally two days ago that acknowledged risks. There’s also still a sizeable focus on safety and people with roles dedicated to it at open ai