Anthropic turns Claude Code’s auto mode on by default: here’s what’s changing for users
Anthropic has said that it is making Auto mode the default mode for all Pro, Max and Team users starting from 14th August. Notably, the feature had been announced for Claude Code users back in March and allows the AI tool to make decisions on behalf of users.
Anthropic’s Head of Claude Code, Boris Cherny, wrote in a post on X (formerly Twitter), “The team and I use Auto mode exclusively, and have been for many months. I couldn’t imagine going back to permission prompts! Really excited to get this out to everyone.”
What is changing in Claude Code?
Anthropic says that Claude Code will run Auto mode by default for Pro, Max and Team users.
Before the change takes place, Anthropic says it will send a one-time in-app notification if the user has already set a default permission mode. Meanwhile, the company adds that users will continue to have the choice of changing back their default mode by clicking on or pressing the Shift+Tab shortcut.
“Auto mode also lets Claude work autonomously for longer stretches. This makes models built for long-running work, like Claude Opus 5, more practical to leave running for hours on large tasks,” Anthropic said in a blog.
The AI startup said that Auto mode will be opt-in for Claude Enterprise users for now, but it will also be made default for these users in the coming months.
Why is Anthropic shifting to Auto Mode?
Anthropic gave two reasons for shifting to Auto Mode on Claude Code. First, Anthropic says that Auto mode ‘matched or beat manual permission review on every safety measure we tracked.’
In fact, Anthropic says in a study of 1,053 paid testers it ran, human testers identified and rejected dangerous commands just 13.6% of the time, while Auto mode blocked 89% of the same commands. Moreover, the company says that Auto mode blocked 800 dangerous commands that human testers approved, while humans blocked only six commands that Auto mode allowed.
Second, the company says that Auto mode makes long-horizon tasks much more viable by allowing Claude to run uninterrupted in the background for hours.
Anthropic also revealed that Auto mode has already been the default for Claude Code internally at the company and has helped prevent many harmful actions.
In one case, Claude attempted to publish a report to a public code-sharing website after an upload to an internal drive failed. Auto mode blocked the action and asked the user what to do next.
In another incident, Claude attempted to kill around 2,000 pods and queue deletions that could have removed hundreds of GPUs running training jobs. The classifier blocked the destructive commands.
