Anthropic is changing how its coding agent, Claude Code, behaves out of the box. Starting August 14, Auto Mode will be the default setting for Pro, Max, and Team subscribers, according to reports from Firstpost and MSN.
Until now, Claude Code has largely operated on a permission-per-action basis: the agent proposes a step, the developer approves it, and the work proceeds. Under Auto Mode, that constant back-and-forth goes away. As India Today describes it, the agent "will no longer ask for your permission for every task" and instead makes decisions on its own.
The counterintuitive part is Anthropic's justification. The company frames the change as a safety improvement, not a loosening of controls. Per Firstpost, the system can identify and block risky actions while allowing routine ones through. The-decoder.com frames the goal as protecting developers from bad approvals, and eGamers.io reports that Anthropic is specifically targeting "rubber-stamped approvals."
That's the logic worth pausing on. When software asks a person to confirm every action, the person stops reading the prompts. Approval fatigue turns a safety check into a reflex, and a reflexive yes protects no one. Anthropic's bet is that a system which intervenes on the genuinely dangerous handful of actions catches more than a human who has clicked "approve" four hundred times that afternoon.
It is also a meaningful shift in who holds the wheel. As finance.biggo.com puts it, Auto Mode replaces constant human approval — moving the developer from gatekeeper on every step to supervisor of the overall result.
The details available so far are limited: the reports describe the date, the affected tiers, and the broad rationale, but not the full mechanics of how risky actions get classified.
Why it matters: A major AI company is arguing that less human oversight can mean more safety — and shipping that argument as the default to paying customers.