An 89 percent success rate in catching dangerous commands drove Anthropic to make Auto Mode the default for Pro, Max, and Team plans. Human reviewers caught only 13.6 percent of those same risks. This change shifts the developer's role from manual approval to high-level monitoring. It prioritizes automated safety over human oversight in Claude Code.