Humans missed 1 in 3 threats when approving AI coding agent commands
Humans missed 1 in 3 threats approving AI agent commands across 40k game runs
Scale X built a browser game where humans approve or deny commands from an AI coding agent. Across 40k+ runs and 409k decisions, players missed 33.7% of threats on average. The most-missed command was npm run analyze (64.7% miss rate)—it looks routine but exfiltrates data via a script in package.json. Threat miss rates climbed toward the end of sessions, consistent with permission fatigue. Over-blocking was also common: npm config set registry (a safe internal mirror) was blocked 59% of the time.
Why it matters: Scale X logged 409,000 approval decisions in a browser game, and players under time pressure missed 33.7% of threats on average, including 64.7% of npm run analyze cases.