Anthropic is turning Claude Code’s auto mode on by default
TechCrunch Anthony Ha ● Covered by 39 sources
Claude Code’s auto mode is becoming the default for paid users on August 14. Anthropic says it’s safer than manual prompts, which people approve 97% of the time.
Based on reporting by TechCrunch, Anthony Ha — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Anthropic is about to make Claude Code a lot less needy. Starting August 14, auto mode will be the default for Pro, Max, and Team accounts, so the assistant will keep moving without asking for approval at every step unless it decides an action is irreversible, destructive, or outside the user’s environment.
The company first showed off auto mode in March as a way to balance speed and control. Now it’s pushing that idea into the default path, which is usually where a feature stops being a curiosity and starts becoming the product.
Anthropic is also making the safety case for the change. In testing with 1,053 paid users, it says auto mode caught 89% of harmful actions, while human review caught 13.6%. That’s a sharp result, and it cuts against the instinct that more permission prompts automatically mean more safety.
There’s another awkward number in the announcement: users approve 97% of permission prompts in Claude Code. If that’s the habit, then a lot of “oversight” may have been more theater than control. Anthropic says it has been adding protections too, including prompt injection screening and customizable hard deny rules aimed at stopping data exfiltration.
Boris Cherny, who heads Claude Code, said his team has used Auto mode exclusively for months and can’t imagine going back to prompts. That’s a strong endorsement, and a very software-person sentence: if the machine is already doing the work, why keep tapping Yes over and over?
My take — AI-written commentary, not fact-checked reporting
This is the open-secret of agentic AI: the permission prompt is mostly a ritual, not a safeguard. Anthropic is admitting that with a grin and a confidence interval. The real question now is whether users want control, or just the comforting sound of being asked before the machine does what it was going to do anyway.
Read more about this at: TechCrunch