Anthropic's Claude AI Can Now Control Your Computer Like a Human Assistant
In a groundbreaking announcement from San Francisco, Anthropic has revealed that its AI agent, Claude, can now take direct control of your computer to complete tasks autonomously. This new function, dubbed Cowork, enables users to message Claude from their smartphones and assign it various jobs, effectively acting as a digital coworker that operates seamlessly across devices.
How Cowork Transforms Daily Tasks
Claude is designed to handle a wide array of computer-based activities that mimic human actions. According to Anthropic, the AI can open applications, sift through files, populate spreadsheets, and perform other desk-bound duties. In a demonstration video, Claude accessed a user's photo library to locate images of their shop, subsequently cropping them and adding watermarks—all without manual intervention.
The company elaborated on X, stating that Claude delivers outcomes such as spreadsheets, memos, comparison tables, or pull requests directly to the user, rather than displaying every step of the process. This allows individuals to assign tasks from their phones, shift focus to other matters, and return to completed work on their computers. For instance, users can instruct Claude to scan emails each morning or generate reports every Friday, with the AI managing these routines independently.
Cross-Device Functionality and Permissions
Claude operates by gaining permission to access a user's web browser, calendar, messaging platforms like Slack, and other tools. Importantly, all tasks are executed on the computer itself, not the phone, creating an ongoing conversation that spans two screens. Users receive push notifications on their phones when tasks are finished or if Claude requires approval to proceed, ensuring oversight and interaction.
Potential Risks and Cybersecurity Warnings
Despite its capabilities, Anthropic cautions that Claude may encounter errors or malicious content during task execution. The company warns that manipulated instructions, unexpected commands, or phishing links opened in the browser could lead to irreversible actions. This highlights significant cybersecurity concerns, as the AI's autonomous nature introduces vulnerabilities that users must navigate carefully.
Industry Context and Expert Insights
Agentic AI, which acts independently unlike traditional chatbots, is gaining traction in Silicon Valley. Applications range from booking flights to social interactions on platforms like Moltbook. However, tech experts express reservations about the rapid evolution of this technology. Mark McClain, CEO of SailPoint, notes that while these agents are not akin to dystopian entities like the Terminator, they pose privacy risks by storing and potentially sharing user data in opaque ways.
McClain emphasizes the temptation to treat AI agents as confidants, given their helpful advice and seemingly safe platforms. Yet, unlike human friends, agents might disseminate information without user visibility or control. Supporting this concern, a Microsoft report shared exclusively with Metro identifies 'shadow AI agents'—unregulated software with unclear data practices—as a major cybersecurity issue for businesses.
Availability and Future Developments
Anthropic is rolling out Cowork initially to Max users, with plans to extend access to Pro subscribers subsequently. As AI continues to integrate into daily workflows, this innovation underscores the balance between enhanced productivity and the imperative for robust security measures in an increasingly automated world.
