Claude Code: Auto Mode Becomes Default Permission Mode on Aug 14

Anthropic is changing how Claude Code handles permissions. Starting August 14, auto mode will be the default permission mode for Pro, Max, and Team users. Auto mode uses a separate classifier to review shell commands and actions before execution, and in Anthropic's testing, it caught dangerous commands far more reliably than manual approval.

Key Details

According to the announcement on X, auto mode's classifier blocked 89% of dangerous commands in testing. In a separate study with 1,053 paid testers, humans manually approving a permission prompt only caught 13.6% of clearly dangerous commands — and that dropped to about 5% after 50 prompts. Anthropic says auto mode matched or beat manual review on every safety measure they tracked.

The change is aimed at enabling long-horizon work. With auto mode, Claude can run for hours between interruptions, which pairs well with models like Opus 5 that excel at multi-hour tasks.

Token Overhead Removed

Auto mode runs a classifier on every tool call, which consumes a small number of extra tokens. As of today, that overhead no longer counts toward usage limits on Pro, Max, and Team plans. The change will eventually roll out to Enterprise plans and API users once auto mode becomes the default for those tiers.

What Users Can Do

• If you've already set a default permission mode, Claude will ask before changing anything.
• You can switch permission modes anytime with Shift+Tab.
• Admins can pin defaultMode or disable auto mode entirely in managed settings.
• You'll get an in-app notification when the default flips.

Why It Matters

This is a significant shift in how AI coding agents handle trust. The data suggests that automated classification can outperform human vigilance in repetitive review tasks, and removing the token overhead removes a major incentive to stick with manual mode. If you're using Claude Code, you'll want to check your settings before August 14 to ensure the change aligns with your workflow.