Anthropic has announced that the auto mode feature in Claude Code will become the default setting for Pro, Max, and Team accounts starting August 14. This update shifts the coding assistant toward greater autonomy, significantly reducing the frequency of manual permission prompts required from human developers during tasks.
The company states that in auto mode, the assistant proceeds independently unless an action is flagged as irreversible or destructive. Internal testing revealed that automated checks caught 89% of harmful actions, outperforming human reviewers who successfully intercepted only 13.6%, often due to habitual approval fatigue. To mitigate risks, Anthropic is also introducing enhanced safeguards, including prompt injection screening and customizable hard-deny rules.
- Auto mode becomes the default for Pro, Max, and Team tiers starting August 14.
- The assistant executes tasks independently with fewer manual permission prompts.
- Testing showed automated oversight caught significantly more harmful actions than humans.
- New security measures include prompt injection screening and deny rules.
Sources:
