Anthropic is turning Claude Code’s auto mode on by default
Claude Code will start acting without step-by-step approval for Pro, Max, and Team users on August 14.
Anthropic says auto mode proceeds unless an action is irreversible, destructive, or outside the user’s environment. The company says testing with 1,053 paid users found auto mode caught 89% of harmful actions, compared with 13.6% for human review. It is also adding prompt injection screening and customizable hard-deny rules aimed at blocking risks like data exfiltration. TechCrunch AI's note
Anthropic says auto mode proceeds unless an action is irreversible, destructive, or outside the user’s environment. The company says testing with 1,053 paid users found auto mode caught 89% of harmful actions, compared with 13.6% for human review. It is also adding prompt injection screening and customizable hard-deny rules aimed at blocking risks like data exfiltration. TechCrunch AI's note
score 5