Subscribe
AI Coding
3 min read

Anthropic Makes Auto Mode Default in Claude Code for Pro, Max, and Team Plans

Starting August 14, 2026, new Claude Code sessions on Pro, Max and Team plans will run in auto mode by default, Anthropic announced on its blog. Users who have pinned a different default keep it, and users who set a different default may get a one-time prompt asking whether to switch. What Auto Mode Does Auto Mode routes every tool call through a classifier that screens for irreversible, destructive, or externally-aimed actions. When the classifier blocks something, Claude attempts a safer al

Anthropic Makes Auto Mode Default in Claude Code for Pro, Max, and Team Plans

Starting August 14, 2026, new Claude Code sessions on Pro, Max and Team plans will run in auto mode by default, Anthropic announced on its blog. Users who have pinned a different default keep it, and users who set a different default may get a one-time prompt asking whether to switch.

What Auto Mode Does

Terminal showing AI classifier decision on command safety

Auto Mode routes every tool call through a classifier that screens for irreversible, destructive, or externally-aimed actions. When the classifier blocks something, Claude attempts a safer alternative or asks for permission. After repeated blocks, the session falls back to manual approval.

The classifier uses a small number of extra tokens per tool call, and Anthropic no longer charges Pro, Max and Team users for that overhead. Auto mode remains opt-in for now on Claude Enterprise, the Claude API and the cloud platforms; Anthropic plans to make it the default there within about a month (Anthropic).

Safety Data Behind the Switch

Anthropic cites a controlled study, incidents from its own internal use, and an outside prompt-injection evaluation:

1. Controlled study (1,053 paid testers): human reviewers caught 13.6% of dangerous commands, against 89% for auto mode. Humans blocked about 17% early in a session, falling to about 5% after 50 or more prior prompts.

2. Internal use: auto mode stopped Claude from publishing a report to a public code-sharing site after an internal upload failed, and flagged a mass process kill across roughly two thousand pods that would have torn down in-flight GPU training jobs (Anthropic).

3. Independent audit by Trajectory Labs: 72 prompt-injection scenarios × 10 runs each = 720 attempts against Claude Fable 5, Opus 5, and Sonnet 5 in Auto Mode. Zero succeeded. For comparison, OpenAI's GPT-5.6 Sol in Codex Auto-Review mode allowed 5.83% of attacks through.

Productivity Impact

Among Team and Enterprise adopters, auto mode users ship about 25% more pull requests, according to Anthropic. The company says fewer interruptions make long-running agent work more practical.

Caveats

Anthropic acknowledges that the classifier does not eliminate risk. "For high-stakes changes to production infrastructure, we still recommend reviewing Claude's actions yourself," the blog post states.

Update, November 9, 2026: After Apiiro found passwords and keys saved by Claude Code's "Yes, and don't ask again" button in public GitHub files, Boris Cherny of the Claude Code team told Hacker News readers that everyone should use auto mode instead of permission rules. A commenter suggested auto mode could still save such rules when it asks about sensitive commands, and Cherny called it a good catch.

Update (Oct 19, 2026)

  • When Claude cannot make progress, auto mode falls back to manual approvals after three blocks in a row or twenty across a session (Anthropic).
  • Classifier overhead is not charged on Pro, Max and Team plans. Auto mode stays opt-in on Enterprise, the Claude API, Amazon Bedrock, Google Cloud and Microsoft Foundry for now, and Enterprise admins can make it the default through managed settings (Anthropic).
  • Admins can pin an org-wide default with `defaultMode` or turn auto mode off with `disableAutoMode` (configuration docs).
  • Corrections: Anthropic's internal incident was a blocked mass process kill across roughly two thousand pods, not 2,000 processes that auto mode killed. We also removed product features and an oversight claim that the blog post does not make.

Sources

AI CodingAnthropicClaude CodeDeveloper Tools