Anthropic sets Claude Code to Auto Mode by default to protect developers from bad approvals

Editor's take

Anthropic is now defaulting its Claude Code assistant to an automated safety mode for its paid tiers, a move aimed at preventing the generation of harmful code. This shift leverages an AI classifier, which demonstrated a significantly higher detection rate for dangerous commands (89%) compared to human reviewers (13.6%) in Anthropic's internal testing.

This development matters as it highlights a growing industry focus on AI safety within development tools, particularly as complex LLMs are increasingly integrated into coding workflows. Developers using Claude Code, especially those on Pro, Max, and Team plans, will now have a default layer of protection against accidentally generating or approving insecure or malicious code snippets, a critical concern in software supply chain security.

The immediate next step to observe is the real-world efficacy of this automated classifier outside of controlled tests, and whether similar safety mechanisms will become standard across competing LLM coding assistants like GitHub Copilot or Google's Gemini. The long-term impact will hinge on how this default setting affects developer productivity and the rate of false positives or negatives in diverse coding environments.