AI firm Anthropic has announced that auto mode will be the default setting for their code editor, starting on August 14. This move aims to streamline the coding process by reducing human oversight.

The company first tested this feature in March, highlighting its potential to balance speed and control. In a study with over 1,053 testers, auto mode proved significantly more effective at catching harmful actions than manual review. The findings suggest that users are often too lenient when approving permissions.

In an X post, Boris Cherny, head of Claude Code, stated his team uses auto mode exclusively and has done so for months. He expressed disbelief at the idea of reverting to permission prompts, indicating a growing preference for automated processes within AI communities.

Anthropic is also introducing additional safety measures such as prompt injection screening and customizable hard deny rules to prevent data exfiltration. These steps aim to address concerns about security while embracing automation.