Anthropic Sets Claude Code’s Auto Mode as Default, Citing Improved Safety Over Manual Review

Anthropic announced that Claude Code will switch to auto mode by default for Pro, Max, and Team accounts starting August 14, reducing the level of human oversight required during coding sessions. First introduced as a test feature in March, auto mode allows Claude Code to proceed with tasks without requesting approval at each step, pausing only when an action is deemed irreversible, destructive, or directed outside a user’s own environment.

The company said internal testing found auto mode outperformed manual review in catching harmful actions. In a study involving 1,053 paid testers, auto mode identified 89% of potentially harmful actions, compared to just 13.6% caught through human review, a gap Anthropic attributed partly to users habitually approving 97% of permission prompts rather than scrutinizing each one.

Boris Cherny, who leads Claude Code, said on social media that his team has relied exclusively on auto mode for months and would not want to return to a permission-prompt system.

Anthropic said it has also introduced additional safety measures alongside the update, including prompt injection screening and customizable hard deny rules designed to prevent risks such as unauthorized data exfiltration during automated coding sessions.