Anthropic is turning Claude Code’s auto mode on by default

Programming with Claude Code will soon require even less human oversight.

Anthropic is set to roll out Claude Code’s auto mode as the default for Pro, Max, and Team accounts beginning August 14, according to the company. Originally unveiled as a test in March, the auto mode mode aims to strike a balance between speed and control. When Claude Code operates in auto mode, it will execute actions without prompting human approval, unless the action is deemed "irreversible, destructive, or aimed outside your environment."

During testing with 1,053 paid users, auto mode was found to be safer than manual review. In the study, auto mode detected 89% of harmful actions, compared to just 13.6% caught by human review. The latter was also found to have users approve 97% of permission prompts. Anthropic's Claude Code Head Boris Cherny stated that his team exclusively uses auto mode, and they haven't looked back since.

The company has also added new safety features, including prompt injection screening and customizable hard deny rules, to prevent issues such as data exfiltration.

Written by urgent.news from TechCrunch's reporting — not their text. Machine-written — it may contain errors, so check the original before relying on it.