Claude Code now works on auto mode by default, Anthropic says it is safer this way

Anthropic has changed the way Claude Code works by default. Now the coding agent will no longer ask for your permission for every task. Instead, it will now make decisions by itself in auto mode, unless you change the settings.

In Short

• Anthropic is making auto mode the default for Claude Code
• In this, Claude Code works on its own, doesn't ask for human approval
• Anthropic says it is safer this way

In March this year, Anthropic announced an auto mode for Claude Code. This mode allowed the coding agent to make decisions without your approval in most but not all tasks. And now, Anthropic is making auto mode the default setting for Claude Code. The AI startup says that when the AI agent works on its own, it can flag dangerous tasks better than humans.

On X, Anthropic said Claude Code will switch to auto mode by default for Pro, Max and Team users from August 14.

Claude Code head Boris Cherny wrote on X, "The team and I use Auto mode exclusively, and have been for many months. I couldn't imagine going back to permission prompts!"

To give you some context, in auto mode, Claude Code does not stop for approval at every step. Instead, each command and task is checked by a separate classifier, and the system proceeds unless an action is judged to be irreversible, destructive, or aimed outside the user's environment.

Claude is better at finding risks than humans?

The company said it is making the change because auto mode performed better than manual permission review in its testing. As per Anthropic, auto mode could flag 89 per cent of dangerous commands – actions that may cause harm to the user’s systems – as opposed to only 11 per cent spotted by humans when they manually approved Claude Code.

This gap widened in longer sessions, with human detection falling to about 5 per cent after 50 or more prior prompts, while auto mode remained steady. The company also said users approve 97 per cent of permission prompts in Claude Code, suggesting that manual review can become habitual.

In case you are unaware, when auto mode finds a dangerous command, it is blocked, and Claude usually tries a safer route or asks the user directly. If it is blocked three times in a row, or 20 times in a session, it falls back to manual approvals.

Apart from safety, Anthropic says that in auto mode, Claude runs longer between interruptions so users can leave multi-hour tasks running in the background without "babysitting permissions.”

The auto mode also works well with models built for long-running tasks, such as Opus 5. The AI startup added that users can start one or several tasks in parallel and return to completed pull requests for review.

According to Anthropic, Teams and Enterprise adopters using auto mode ship about 25 per cent more pull requests. Adobe, Nuro, Gusto and Garner Health already use auto mode as a production default.

Keep in mind that Anthropic says the auto mode reduces risk for most users, but does not remove it completely. The startup recommends that users still review high-stakes changes to production infrastructure themselves.

Who can use auto mode as default?

Auto mode will become the default for Claude Code for Pro, Max and Team users from August 14. Users will receive an in-app notice when the default changes, can switch modes with Shift+Tab, and admins can pin a default mode or disable auto mode entirely.

Anthropic says that the extra tokens used by the auto mode classifier will no longer count towards usage limits from now.

It will remain opt-in on Claude Enterprise, the Claude API, Claude Platform on AWS, Amazon Bedrock, Google Cloud's Agent Platform and Microsoft Foundry. Though Anthropic said it plans to make it the default there in the coming month after notifying administrators.

- Ends