Anthropic hands Claude Code more control, but keeps it on a leash

Anthropic introduced an 'auto mode' research preview for Claude Code that lets the AI autonomously decide which developer actions are safe to execute without human approval, using built-in safeguards to check for risky behavior and prompt injection. The feature builds on the existing 'dangerously-skip-permissions' command by adding a safety layer, though Anthropic has not detailed the specific criteria used to distinguish safe from risky actions. Auto mode will roll out to Enterprise and API users and currently supports Claude Sonnet 4.6 and Opus 4.6.