Block the Prompt, Not the Work: The End of "Doctor No"

This article argues that traditional endpoint and network controls are driving employees to unsafe workarounds (personal email, unmanaged AI tools, and browser extensions), leaving a blind spot for sensitive data exfiltration; it advocates for session-level, agentless governance (prompt-level DLP, extension risk scoring, clipboard/upload controls) as the modern approach and illustrates the problem with an anecdote of a law firm whose users installed an AI wrapper extension that routed corporate traffic through overseas servers.