Claude Code makes auto mode the default Aug 14 for Pro/Max/Team
Approval prompts vanish for most sessions. Anthropic's classifier blocked 89% of risky commands to humans' 13.6%, and it's dropping the overhead fee.
Copy markdown
The default flips on August 14
For Pro, Max, and Team plans, new Claude Code sessions will start in auto mode: Claude runs tool calls without asking at each step, pausing only for irreversible, destructive, or out-of-your-environment actions. Enterprise stays opt-in for now.
Why: you rubber-stamp your own prompts
Anthropic says users approve 97% of permission prompts, so in a 1,053-tester study humans caught just 13.6% of dangerous commands versus 89% for the classifier. Auto-mode sessions logged unintended harm 2.4% of the time against 6.3% for manual approval.
The safety tax is now free
The classifier overhead that inspects every tool call is no longer billed on Pro, Max, and Team plans, effective immediately. The guardrail stops eating into your token budget, so leaving it on costs you nothing.
New guardrails ship alongside it
Hard deny rules block data exfiltration, a git-status check gates destructive commands, and prompt-injection screening covers externally-sourced content. After hardening against Apollo Research's synthetic attacks the miss rate fell to 7% from 12%, and none of 720 injection attempts got through.
Before Aug 14: pin your setting
Auto mode won't override a permission-prompt default you or your admin has already pinned, and existing users get a one-time switch prompt. Anthropic reports teams on auto mode ship ~25% more PRs; Claude Code lead Boris Cherny says he's used it exclusively for months.