Claude Code makes auto mode the default with new security guarantees

Anthropic set auto mode as the default configuration in Claude Code for Pro, Max, and Team plans starting August 14th. The company released security evaluations showing that automatic mode would block 89% of dangerous actions compared to just 13.6% of refusals by human reviewers. In indirect prompt injection tests with 720 attack attempts, none succeeded against Claude Fable 5, Opus 5, or Sonnet 5 models. Despite positive results, experts raise questions about uncovered vulnerabilities, such as malicious third-party packages that could exfiltrate data during development processes.