Anthropic says it will change the default behavior of Claude Code on August 14 by enabling auto mode for new sessions on its Pro, Max, and Team plans. In auto mode, Claude executes eligible coding actions without repeatedly asking for permission, while a system evaluates tool calls and blocks actions deemed irreversible or harmful. Anthropic also says some existing users may be prompted to switch if they previously chose a different default.
The outlets frame the change primarily as a safety and workflow update. Help Net Security and others cite Anthropic’s testing results, where human reviewers reportedly caught a smaller share of dangerous commands than the automated approach. NDTV adds that Anthropic builds in additional safeguards aimed at issues such as prompt injection, data exfiltration, destructive Git commands, and access to sensitive data. Times of India similarly describes auto mode as generally safer than manual checks and positioned as improving programming efficiency.
Anthropic is also described as keeping auto mode optional for some enterprise offerings, with the default change targeted to Pro, Max, and Team.
