Claude, the AI from Anthropic, can now take control of your computer to help complete tasks. If you subscribe to a qualifying plan, you can take advantage of this feature.
Earlier this year, the open-source framework OpenClaw gained a lot of attention. It allowed developers to create AI tools that can perform simple commands on a computer. Recently, Nvidia launched NemoClaw, a framework that simplifies working with OpenClaw, including security options.
Claude connects with apps like Google Calendar and Slack to find the right tools for the job. If a connection isn’t available, it can still operate as if it were using a keyboard and mouse. It can open files, browse the web, and execute commands, all while checking in with you for permission before taking action. You can stop it anytime if you feel uneasy.
However, AI that operates autonomously can pose risks. Experts warn that such systems can act quickly, sometimes unexpectedly. There’s a concern that malicious actors could hijack these AI tools, putting your personal data at risk. To address these issues, Anthropic has built in safeguards to reduce vulnerabilities, like prompt injections.
Users need to be cautious. Although the feature is designed to be safe, it’s still new and has potential errors. Anthropic advises against using it with sensitive applications, and some of these apps are disabled by default.
Currently, the feature is accessible for Claude Pro and Claude Max subscribers on MacOS. It pairs well with a tool called Dispatch, letting you assign tasks to Claude via your phone. For example, it can check your email or prepare a morning briefing while you’re away.
The combination of these features is exciting, but they are still in the early stages. Some tasks might not go as planned initially. By releasing this preview, Anthropic aims to gather feedback and improve the tool.
To stay informed on AI and security frameworks, you might find this recent report on AI risks insightful.
