Full story
Auto Mode Becomes Default
Anthropic said Claude Code will switch to auto mode by default for Pro, Max and Team users from August 14, replacing the need for step-by-step permission prompts on most tasks.
“from August 14”
In auto mode, the coding agent proceeds unless an action is judged to be “irreversible, destructive, or aimed outside the user's environment,” and it can fall back to manual approvals after three consecutive blocks or 20 blocks during a session.
Anthropic also said auto mode is designed to flag dangerous commands better than humans, with the company claiming it could flag 89 per cent of dangerous commands versus only 11 per cent spotted by humans when they manually approved Claude Code.
The company said its testing found auto mode performed better than manual permission review, and it added that users approve 97 per cent of permission prompts in Claude Code.
Anthropic said the change is intended to reduce interruptions so users can leave multi-hour tasks running in the background without “babysitting permissions.”
How Anthropic Says It’s Safer
Anthropic said Claude Code’s auto mode reviews shell commands and actions with a separate classifier, and that the system is designed to block actions considered irreversible or destructive.
In testing described by Tech in Asia, auto mode caught 89% of harmful actions while human review only caught 13.6%, and the company linked the gap to the idea that “manual review can become habitual.”
Gadgets 360 reported that Anthropic stopped charging users on select plans for the additional classifier tokens required for auto mode, and it said the classifier routes each tool call through a classifier instead of showing permission prompts before individual actions.
Gadgets 360 also said Anthropic added safeguards covering prompt injection, data exfiltration, destructive Git commands, and sensitive data access, and it described hard-deny rules for actions such as data exfiltration attempts.
TechCrunch said Anthropic’s announcement described auto mode as proceeding unless an action is determined to be “irreversible, destructive, or aimed outside your environment.”
Rollout, Controls, and Users
Anthropic said 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.
“The team and I use Auto mode exclusively, and have been for many months”
The company said auto mode will remain opt-in on Claude Enterprise, the Claude API, Claude Platform on AWS, Amazon Bedrock, Google Cloud's Agent Platform and Microsoft Foundry, and it plans to make it the default there in the coming month after notifying administrators.
Boris Cherny, head of Claude Code, wrote on X that “The team and I use Auto mode exclusively, and have been for many months,” and he added, “I couldn't imagine going back to permission prompts!”
The Decoder said Anthropic also urged caution for “high-stakes changes to production infrastructure,” recommending that developers still review Claude’s actions themselves.
BigGo Finance reported that in a study involving 1,053 paid testers, auto mode caught 89% of dangerous commands while humans in a manual review scenario spotted only 13.6%, and it said the company attributes the gap to what it calls “manual review fatigue.”
