Anthropic announced that starting August 14, the Claude Code automatic mode will be enabled by default for Pro, Max, and Team accounts, further reducing the need for manual intervention in AI programming. This mode aims to improve code development efficiency while balancing automated execution capabilities with risk control through security mechanisms.

Anthropic first introduced a test version of the Claude Code automatic mode in March of this year, positioning it as a new interaction method that balances speed and control. In the automatic mode, Claude Code will no longer request user confirmation before each step, but will directly execute tasks when the system determines that the operation does not involve "irreversible, destructive, or external environment-oriented" risks.

claude, claud

Anthropic stated that test results showed the automatic mode outperformed manual review in terms of security detection. In a study involving 1,053 paid testers, the automatic mode identified 89% of harmful behaviors, while manual review detected only 13.6%. The company noted that manual review might lead to habitual approvals due to long-term use, with users approving 97% of Claude Code's permission requests during the test.

Boris Cherny, head of Claude Code, said his team has been using the automatic mode for a long time and stated they can no longer return to a development approach reliant on permission prompts. At the same time, Anthropic continues to improve its security system, including adding a fast injection filtering mechanism and customizable hard rejection rules to reduce risks such as data leaks.

As AI programming assistants gradually evolve toward agent-like capabilities, the balance between automated execution and security constraints has become a key focus for the industry. This adjustment by Claude Code also reflects the evolution of AI development tools from assisting programming to higher autonomy in task execution.