Anthropic has activated auto mode by default for paid users of Claude Code, enabling the AI to execute routine tasks with minimal permissions, signalling a shift towards more autonomous AI workflows amid safety concerns.
Anthropic is making its Claude Code assistant more autonomous by turning on auto mode by default for paying users on Pro, Max and Team plans. The change means the system can decide for itself whether to carry out many routine actions, rather than stopping repeatedly for permission, although it should still pause for anything it judges to be irreversible, destructive or outside the user’s environment.
The move is aimed at reducing the friction that can slow software development when an AI tool is asked to work for long stretches. Anthropic says Claude can now continue for hours on large tasks with fewer interruptions, which it argues should make the product more useful in coding and engineering workflows. In practice, that shifts the system closer to an agentic model, where the software not only suggests work but also executes more of it on the user’s behalf.
That autonomy comes with clear risks. Concern around so-called rogue AI systems has grown after reports that some models have been involved in hacking activity or other harmful behaviour, including cases involving Claude itself. Anthropic says it has spent the past several months testing whether auto mode performs better than manual approval, including trials with 1,053 paid testers, and says the feature matched or outperformed manual review on every measure it examined. The company also says repeated blocks should trigger a return to manual approval.
The update is limited for now. According to Anthropic, auto mode is switched on by default only within Claude Code, not in the standard chatbot interface, and only for users on the relevant subscription tiers. TechRadar reported that the feature is being treated as a research preview, with rollout initially focused on team customers and wider support expected later, while Anthropic’s own documentation says the mode can be configured around trusted repositories, buckets and domains. Boris Cherny, Anthropic’s head of Claude Code, said on X that his team uses auto mode exclusively and has done so for months, adding: “I couldn’t imagine going back to permission prompts!”
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
