Anthropic is turning Claude Code’s auto mode on by default | TechCrunch

Anthropic has announced that its terminal-based coding assistant, Claude Code, will transition to utilizing "auto mode" as the default setting for paid users, including those on Pro, Max, and Team tiers, starting on August 14.

Initially introduced in an experimental capacity earlier this spring, the feature is designed to increase developer productivity by minimizing repetitive confirmation steps. Under this new setup, the artificial intelligence agent will execute multiple programming steps autonomously without requiring human intervention for every prompt, only stopping its workflow if an action is deemed highly risky, irreversible, or attempts to access external networks beyond the local environment.

In addition to efficiency gains, Anthropic claims that auto mode significantly enhances security. The firm's research involving over one thousand premium testers revealed that the automated system blocked nearly ninety percent of harmful actions, whereas human operators caught less than fifteen percent, largely because manual reviewers frequently fall into a habit of approving prompts blindly. To address potential vulnerabilities, the organization is also rolling out extra defensive layers for the software, including screening mechanisms for prompt injection and customizable prevention rules to block unauthorized data exfiltration.