Anthropic Makes Claude Code Auto Mode the Default for Pro, Max and Team Users

Anthropic is making Claude Code’s auto mode the default for Pro, Max and Team accounts starting August 14, allowing its AI coding agent to complete more tasks without repeatedly asking users for permission.

Anthropic introduced a test version of auto mode in March as a way to balance faster autonomous coding with safety controls. Instead of requesting human approval at every step, Claude Code can proceed automatically unless it determines that an action is irreversible, destructive or directed outside the user’s environment.

Anthropic says auto mode can be safer

The company says its testing found auto mode was more effective at identifying potentially harmful actions than relying on users to review permission requests manually.

In a study involving 1,053 paid testers, auto mode detected 89% of harmful actions, compared with 13.6% detected through human review. Anthropic noted that users approved 97% of Claude Code permission prompts, suggesting repeated requests can make manual review less effective over time.

Claude Code head Boris Cherny said on X that his team has used auto mode exclusively for months and that he would not want to return to repeated permission prompts.

More safeguards for autonomous coding

Anthropic is also adding security measures as Claude Code becomes more autonomous. These include prompt injection screening designed to detect malicious instructions that could influence the coding agent.

The company is introducing customisable hard deny rules as well, allowing organisations to block specific actions regardless of what Claude Code decides during a task. Anthropic said these controls can be used to prevent risks such as unauthorised data exfiltration.

Making auto mode the default reflects a broader shift toward coding agents that can independently plan and execute longer workflows while requiring human intervention mainly for higher-risk actions.