Claude from Anthropic Challenges ChatGPT by Controlling Your Computer

Claude Takes Over Your Computer

Anthropic, the company behind Claude, continues to push the boundaries of its artificial intelligence to make it an unparalleled productivity tool. Claude stands out particularly with its AI assistant, Claude Cowork, which is designed to perform tasks on behalf of the user, thereby freeing them up for more strategic activities. Claude is positioned as a direct competitor to ChatGPT, aiming to offer more advanced and integrated features.

So far, Claude Cowork and Claude Code, a version dedicated to developers, relied on connectors to interact with various services such as Slack, or took control of Google Chrome via an extension. However, some tasks remained beyond the reach of these connectors and could not be performed in a simple browser. To address this limitation, Anthropic has introduced a groundbreaking feature that allows its AI to take full control of the user's computer.

Anthropic explains that "when Claude does not have the necessary connector or tool, it navigates directly on your screen: it clicks, types text, and opens applications just as you would yourself. It can operate in your browser, open files, and automatically launch your development tools, without prior configuration." Furthermore, this feature can be controlled remotely via a smartphone.

A Last Resort Solution

Currently, this new capability is available on macOS, with a Windows extension planned for the near future. To take advantage of this innovation, users of Claude Cowork or Claude Code must subscribe to a paid plan. Anthropic specifies that Claude will always prioritize more direct methods, via connectors, to perform an action for the user. Only in the absence of a connector will Claude first use a browser, and as a last resort, it will take control of the cursor and keyboard to manipulate the computer.

However, this approach, while functional, is slower than using connectors. Additionally, the AI's navigation of the operating system carries risks. For example, Anthropic emphasizes that to use this interface, the artificial intelligence relies on screenshots, which can lead to the inadvertent capture of sensitive information when Claude takes control of a machine.

Risks Associated with Sensitive Applications

To mitigate these risks, Anthropic has implemented several security measures. For instance, Claude systematically asks for the user's permission before opening an application on the computer. It is also possible to create a list of applications that Claude should not use. Furthermore, Anthropic has issued several recommendations for users:

• "Do not grant permission to access sensitive applications (such as banking, health, or government applications)," advises the developer.

On the other hand, an application developer can ask Claude to open a smartphone emulator and navigate the interface of a developing application to identify malfunctions. This feature also allows an employee to extract data from internal tools of a company that do not have connectors. Thus, Claude's AI agent can automate even more tasks by taking control of the computer when the necessary connectors to accomplish a task do not exist.