Claude Code Auto Mode: Safer AI Coding by Default

Anthropic is making Claude Code's auto mode the default setting for new sessions across its Pro, Max, and Team plans starting August 14, 2026. This change replaces constant permission prompts with an automated system. The company says this shift is based on data showing auto mode is safer. In a test, auto mode blocked 89% of dangerous commands, while human testers only refused 13.6% of the time. Independent evaluations also found that auto mode successfully defended against 720 indirect prompt injection attacks. Anthropic will also stop charging for the extra tokens used by the auto mode classifier. While the company highlights productivity gains, some analysts still point to potential risks, such as malicious third-party packages. The bottom line is this change aims to improve both safety and efficiency for developers using Claude Code.