AI in defense

OpenAI writes the Pentagon's looser human-control standard into its own policy

"Appropriate human judgment" does not oblige a person to approve each lethal strike. With no agreed international standard to cite, the human-in-the-loop bar is now being set in vendor usage policies.