Anthropic announced it will activate auto mode as the default setting for Claude Code sessions on Pro, Max, and Team subscription plans beginning August 14, 2025. Users who previously selected a different default mode will receive a one-time prompt asking if they want to switch to the new auto mode setting. Enterprise plan customers will continue to have the option to select their preferred mode.
The change follows a controlled experiment involving 1,053 paid professional testers that revealed significant differences in security effectiveness between human review and automated review. The study found that human reviewers caught only 13.6% of dangerous commands during code execution, while auto mode successfully identified 89% of such threats.
Auto mode in Claude Code automatically reviews and evaluates potentially risky commands before execution, providing an additional layer of security for code generation and execution tasks. The feature analyzes commands for security risks and can prevent harmful operations from running without requiring manual human intervention for each action.
The testing results suggest that relying solely on human review for code security may leave significant gaps in protection, particularly as AI-generated code becomes more complex and widespread. The nearly sixfold difference in detection rates between automated and human review highlights the challenges users face in manually identifying all potential security risks during development workflows.
Organizations using Claude Code on affected plans should prepare for the transition to auto mode as the default setting. Teams may want to review their current workflows and security policies to ensure they align with the new default configuration. Users who prefer manual review can still opt out of auto mode, though the experimental data suggests this may reduce their security posture significantly.
Source: https://www.helpnetsecurity.com/2026/08/10/anthropic-claude-code-auto-mode/