Anthropic has introduced a research preview of a computer-use feature for its AI, specifically within the Claude Cowork and Claude Code applications. This functionality allows the model to interact directly with a user’s computer by controlling the screen, keyboard, and browser. Currently limited to macOS, the feature requires the desktop application to be active and relies on explicit user permissions for each action to maintain security. Through integration with a tool called Dispatch, users can assign tasks from mobile devices and have the AI execute them on their desktops, streamlining workflows like checking emails or generating reports. While the system prefers using official service connectors for platforms like Slack or Google Calendar, it is capable of manual navigation when necessary.
Beyond these technical advancements, the company is navigating significant challenges. The platform recently dealt with technical disruptions that impacted user stability. Furthermore, Anthropic is embroiled in a legal conflict with the U.S. government regarding the military use of its systems. This dispute involves the defense department’s desire for unrestricted AI use, leading to a lawsuit over national security concerns and the ethical boundaries of AI in defense. As developers continue to refine the computer-use feature based on feedback, they advise users to avoid processing sensitive data during this experimental phase. The goal remains to evolve the AI into a more capable digital collaborator despite the broader political and operational hurdles the organization faces.
