Anthropic makes Claude Code’s auto mode the default for Pro, Max, and Team accounts
Auto mode will proceed unless actions are deemed irreversible, destructive, or outside the user’s environment, starting August 14.
1 source · cross-referenced
- Auto mode will be enabled by default for Anthropic’s Claude Code Pro, Max, and Team accounts beginning August 14, 2026.
- The feature executes actions autonomously unless they are flagged as irreversible, destructive, or outside the user’s environment.
- In a study with 1,053 paid testers, auto mode caught 89% of harmful actions, compared to 13.6% caught by human review.
- Anthropic added prompt injection screening and customizable hard deny rules as additional safeguards.
Anthropic announced it will enable Claude Code’s auto mode by default for Pro, Max, and Team accounts starting August 14, 2026. The company describes auto mode as proceeding with actions unless they are determined to be irreversible, destructive, or aimed outside the user’s environment.
The decision follows internal testing in which auto mode caught 89% of harmful actions in a study with 1,053 paid testers, compared to 13.6% caught by human review. Anthropic attributes the gap in part to habitual approvals in manual review, noting users approve 97% of permission prompts in Claude Code.
To mitigate risks, Anthropic said it has added new safety features such as prompt injection screening and customizable hard deny rules aimed at preventing actions like data exfiltration.
Claude Code Head Boris Cherny stated in a post on X that he and his team have used auto mode exclusively for months and would not return to permission prompts.
- Aug 10, 2026 · TechCrunch — AI
AI-focused hedge fund invests $400M in chip startup Source Foundry
Trust75 - Aug 10, 2026 · Simon Willison’s Weblog
GitHub Models retired after unified LLM API for GitHub Actions
Trust79 - Aug 9, 2026 · AWS — AI News
AWS adds native vector search to DynamoDB with single-digit millisecond latency
Trust84