Claude Code Goes Autonomous

Anthropic is taking a significant step towards reducing human oversight in its AI programming platform, Claude Code. As of August 14, auto mode will become the default for Pro, Max, and Team accounts, replacing the need for manual review at each step.

When in auto mode, Claude Code will proceed unless an action is deemed "irreversible, destructive, or aimed outside your environment." This change is designed to provide a balance between speed and control.

In testing, auto mode proved to be safer than manual review, catching 89% of harmful actions compared to human review's 13.6%.

The company also pointed out that manual review can become habitual, with users approving 97% of permission prompts in Claude Code.

According to Claude Code Head Boris Cherny, "The team and I use Auto mode exclusively, and have been for many months. I couldn't imagine going back to permission prompts!"

To further enhance safety, Anthropic has added new features such as prompt injection screening and customizable hard deny rules to prevent data exfiltration and other potential issues.

This move is expected to improve the overall efficiency and security of Claude Code, allowing users to focus on their work while minimizing the risk of errors or malicious actions.