Back to all articles
•Technology

Anthropic's Claude Code to Default Auto Mode for Advanced Accounts

View original source

Anthropic has announced a significant update for its Claude Code programming platform, introducing 'auto mode' as the default for Pro, Max, and Team accounts starting from August 14. Initially tested in March, this feature, according to a Friday announcement, aims to enhance programming speed and control by allowing more autonomous operation.

  • In auto mode, Claude Code bypasses frequent human prompts unless actions are deemed “irreversible, destructive, or outside the user's environment.”

  • Testing showed auto mode is safer, identifying 89% of harmful actions compared to 13.6% by manual review, highlighting a trend where users approve 97% of permission prompts, indicating possible complacency.

  • Boris Cherny, the Head of Claude Code, supports this streamlined approach, citing personal and team application of auto mode for several months.

  • New safety features are being integrated, such as prompt injection screening and customizable denial rules, to protect against data exfiltration and other security concerns.