Anthropic’s Claude Code gets the “most secure” automatic mode.

Anthropic has Fired “Automatic mode” for Claude Codea new tool that allows artificial intelligence to make permissions-level decisions on behalf of users. The company says this feature offers live programmers a safer alternative between constant hand-holding or giving the model dangerous levels of autonomy.

Claude Code is Able to act independently On behalf of users, which is a useful feature but risky because it can also do things that users don’t want, such as deleting files, sending sensitive data, executing malicious code or… Hidden instructions. Automatic mode is designed to prevent this, flagging and blocking potentially risky actions before they are triggered and giving the agent a chance to try again or ask the user to intervene.

Currently, Automatic mode is only available as a research preview for Team plan users. Anthropic says access will expand to Enterprise and API users in the “coming days.”

Anthropic warns that the tool is experimental and “doesn’t eliminate” the risks completely, and recommends developers use it in “isolated environments.”