Anthropic has announced that its Claude Code and Claude Cowork AI assistants can now directly control a user's local computer desktop to complete tasks, including opening files, using browsers, and running developer tools. This 'computer use' feature, available in a research preview for MacOS subscribers, allows the AI to scroll, click, and navigate the screen when direct API connections aren't available, though it is slower and more error-prone. The capability, which can also be managed remotely via the Dispatch tool, raises significant security concerns about granting AI agents access to sensitive local resources.
Background
The development of AI agents capable of directly interacting with and controlling computer interfaces represents a significant step beyond simple text-based assistants, moving towards autonomous task completion. This trend is part of a broader industry push to create more capable and proactive AI systems that can operate across various software environments.
- Source
- Ars Technica
- Published
- Mar 24, 2026 at 11:45 PM
- Score
- 8.0 / 10