Anthropic has not announced Claude can run computer tasks with no setup

In short: A recent report suggests Claude can ask permission and then use your computer on its own with no setup, but Anthropic’s publicly described features do not match that claim.

What happened

A story published by The Verge said Anthropic updated Claude so it can perform tasks by using your computer for you. The report described actions like opening files, using web browsers and apps, and running developer tools, even when you are away.

Based on Anthropic’s own past announcements, the closest related feature is called “computer use,” introduced in November 2024 as a research preview for Claude 3.5 Sonnet. This feature lets Claude interact with a computer screen by looking at screenshots, then moving the mouse, clicking, and typing, like a remote helper. Anthropic also described this preview as slow and error-prone, and demos have shown it making basic mistakes, like clicking the wrong thing.

Anthropic has also announced local file access for Claude 4 models, where developers can choose to let Claude read certain files on a machine. This is more like giving an assistant a folder of documents to reference, not letting it take over the whole computer. Other Claude tools, like web search, code execution, and coding integrations, are aimed at developer setups and controlled environments, not hands-off desktop control.

Why it matters

Claims about an AI using your computer while you are away raise obvious privacy and safety questions. If the feature is only available as a developer preview with setup and limits, it is a very different situation than a tool that anyone can turn on instantly.

Source: The Verge AI