Anthropic Set Claude Code automatic mode as default to improve programming efficiency

Anthropic announced that from now on, the auto mode of Claude Code will become the default permission setting for Pro, Max and Team users. New sessions use automatic mode unless the user or administrator has pinned other settings. In addition, Anthropic will no longer charge for additional Tokens used for each tool call of the automatic mode classifier. How Automatic Mode Works Automatic mode replaces duplicate approval prompts with classifiers that check each tool call for irreversible, destructive, or out-of-scope operations. If the classifier blocks an operation, Claude will try a more secure method or request permission from the user; if it is continuously blocked, the work stage will return to manual approval mode. Official test data Anthropic stated that its research covered 1,053 paid testers, and the results showed that humans could only detect 13.6% of dangerous instructions, while the automatic mode could intercept 89% of dangerous instructions; after 50 consecutive prompts, human performance dropped to about 5%. Nonetheless, the Anthropic reminder classifier cannot completely eliminate risks, and users are advised to still conduct manual review when making changes to the production environment. Default Mode Impact This change makes long-term operation of the Claude Code more practical and works with routines, voice mode, and in-app Mac browsers. Anthropic pointed out that the number of Pull Requests for Team and Enterprise customers in automatic mode increased by approximately 25%. Automatic mode is launched as the default permission mode starting today. If the user has set other default modes, Claude will ask before making changes; users can still press Shift+Tab to switch modes at any time, or fix a mode with "defaultMode" in the settings.