Claude Code Auto Mode Blocks 89% of Risks vs Human 13.6%

Anthropic enables Auto Mode by default for Pro/Max/Team users on August 14, citing superior safety classification over human oversight which failed to block most dangerous commands.

Woofun AI reports that Anthropic will enable Claude Code's Auto Mode by default for Pro, Max, and Team users starting August 14. The system replaces frequent user confirmations with a safety classifier to determine action execution. Internal testing with 1,053 paid professional testers revealed humans blocked only 13.6% of inserted dangerous commands, compared to 89% blocked by Auto Mode. Human blocking rates dropped to approximately 5% after handling more than 50 permission prompts consecutively. In production, users approve 97% of permission requests, with one-quarter of sessions skipping checks entirely. Auto Mode permits safe operations while blocking high-risk actions like data deletion or external data transmission. Token costs for Auto Mode decisions are waived for Pro, Max, and Team users. Enterprise, API, and major cloud platform users must currently enable the feature manually, with default activation planned for the future.