Anthropic turns Claude Code's auto mode on by default: AI agent autonomy becomes the norm

Anthropic switched Claude Code's auto mode to the default: the coding agent executes a sequence of task steps without asking for confirmation at each one. Previously a user enabled that degree of freedom deliberately — now a deliberate action is needed to restrict it.

Defaults are the most underrated setting in software: most users never change them. When a vendor moves a default toward autonomy, it effectively changes the behavior of thousands of teams at once — more actions happen without a human in the loop, and this is the market's general direction, not one product's quirk. Autonomy itself is not the villain: it is precisely the source of the time savings AI agents are hired for.

The question is what backstops it. A sane frame for a business in Kazakhstan: autonomous steps where a mistake is cheap and reversible, human confirmation where an action touches money, customers or data. In 1Cifer that frame is built into the process itself: agent actions are formalized as tasks, sensitive ones travel an approval route, and every step keeps a history of who allowed what. Running the same audit on your own tools — from the CRM to email auto-replies — takes one meeting.