---
title: "Anthropic Sets Claude Code to Auto Mode Default Following Safety Performance Success"
url: https://projectchintan.com/article/anthropic-claude-code-auto-mode-default-update-2em8b
publisher: Project Chintan
author: Project Chintan Newsroom
section: Technology
published: 2026-08-10T11:37:35.768Z
modified: 2026-08-10T16:01:36.135Z
language: en-IN
---

# Anthropic Sets Claude Code to Auto Mode Default Following Safety Performance Success

Anthropic is transitioning Claude Code to an autonomous default setting for paid users starting August 14. Internal testing revealed that the AI identifies harmful actions at a significantly higher rate than human reviewers.

## Key takeaways

- Claude Code will shift to an autonomous default setting for Pro, Max, and Team users on August 14.
- Internal studies indicate the AI identifies 89% of harmful actions compared to just 13.6% by human reviewers.
- Manual oversight will now be reserved for irreversible or destructive actions rather than every step.
- Anthropic is deploying new safeguards, such as prompt injection screening, to support the autonomous workflow.

## What Happened

Anthropic announced that its AI-driven programming tool, Claude Code, will operate in 'auto mode' by default starting August 14. This update applies to users under Pro, Max, and Team accounts. The feature, which initially entered testing in March, reduces the frequency of human intervention by allowing the AI to execute tasks sequentially without seeking approval for every individual step.

Under this new configuration, the system will only pause for human authorization if a proposed action is classified as irreversible, destructive, or directed outside the user's immediate environment. To support this transition, Anthropic has integrated additional security measures, including screens for prompt injections and customizable 'hard deny' rules designed to block unauthorized data exfiltration.

## Why It Matters

The decision to bypass manual review stems from data suggesting that human oversight is less effective than automated safety checks. In a study involving 1,053 paid testers, Anthropic found that auto mode identified 89% of harmful actions. In contrast, human reviewers caught only 13.6% of the same issues. The company noted that manual review often becomes a habitual action, with users approving 97% of all permission prompts presented to them.

## Key Facts

- The default change takes effect on August 14 for paid account tiers.
- Auto mode was first introduced as a test version in March.
- Internal data shows humans fail to catch approximately 86% of harmful actions that the AI identifies.
- Boris Cherny, Head of Claude Code, stated that the development team has utilized auto mode exclusively for several months.
- New safety protocols include prompt injection screening and specific rules to prevent data exfiltration.

---
Canonical: https://projectchintan.com/article/anthropic-claude-code-auto-mode-default-update-2em8b
Reported from: Multiple Sources