Anthropic’s New Computer‑Use Feature for Claude

Anthropic has added a capability that lets its Claude AI model take control of a computer’s desktop environment. When activated, Claude can see everything displayed on the screen, including personal data, sensitive documents, and private information. This functionality is part of a research preview that aims to explore how AI agents can assist users by performing tasks directly on their machines.

Built‑In Safeguards and Limitations

The company stresses that Claude is equipped with multiple safeguards intended to prevent common risks. These safeguards are designed to block “risky operations” such as moving or investing money, modifying files, scraping facial images, or inputting sensitive data. Anthropic also notes that access to certain “off‑limits” applications—such as investment and trading platforms or cryptocurrency services—will be restricted by default. While these measures are in place, Anthropic acknowledges that they are not perfect and that Claude may occasionally act outside the intended boundaries.

Guidance for Users During the Preview

Anthropic recommends that users begin by granting Claude access only to applications they trust and avoid using the feature with sensitive data. This precaution is intended to mitigate potential exposure of private information while the technology is still being evaluated. The company’s support page reinforces the advice to start with trusted apps and to be cautious about the types of data Claude can see.

Industry Context and Comparable Efforts

The announcement arrives shortly after a wave of similar offerings from other technology firms. Perplexity introduced a Personal Computer feature, Manus launched My Computer, and Nvidia released NemoClaw—each allowing AI agents to control a desktop directly. These developments follow the viral spread of OpenClaw earlier in the year, which demonstrated the appeal of personal AI agents capable of interacting with a user’s computer. In response to that momentum, OpenAI hired the creator of OpenClaw, Peter SteinBerger, to lead the next generation of personal agents.

Implications and Outlook

Anthropic’s rollout reflects a broader industry trend toward integrating AI assistants more tightly with everyday computing tasks. By coupling advanced language models with direct desktop control, companies aim to create more seamless and productive user experiences. At the same time, the emphasis on safeguards highlights ongoing concerns about privacy, security, and the potential for unintended actions by AI systems. As the research preview progresses, Anthropic and other firms will likely continue refining these features to balance utility with safety.

Used: News Factory APP - news discovery and automation - ChatGPT for Business