Anthropic introduces safety guardrails for Claude Code's autonomous mode to prevent unintended file deletion and data breaches.

Anthropic launched 'auto mode' for Claude Code, enabling the AI to make permissions-level decisions independently on users' behalf

The feature addresses the safety risks of AI autonomy by flagging and blocking potentially dangerous actions before execution

Auto mode provides a middle ground between requiring constant user approval and granting the model unrestricted access that could enable file deletion, data leaks, or malicious code execution