Anthropic makes auto mode default for Claude Code Pro, Max and Team accounts

Anthropic will stop asking for user approval at every step of Claude Code. Starting 14 August, auto mode becomes the default setting for Pro, Max and Team accounts.

In auto mode the model runs forward until it detects an action that is irreversible, destructive, or directed outside the user’s workspace. The company first released a beta of auto mode in March, saying it was looking for a balance between execution speed and human oversight. Anthropic now feels comfortable making the setting standard.

The decision is backed by a study involving 1,053 paid evaluators. In Anthropic’s tests, auto mode caught 89% of harmful actions, while a manual human review caught only 13.6%. Users approve 97% of permission requests in Claude Code, so blind approval erodes the value of human control.

Boris Cherny, who leads the Claude Code team, wrote in a post on X: “We have been working only in auto mode for many months and I cannot imagine going back to approval windows.” Anthropic also says it has added extra safety layers, including filters for prompt-injection attempts and custom hard-deny rules designed to block information-leak scenarios.