Microsoft Wants Copilot to Run Your PC—Now It Has to Earn Your Trust
Microsoft Wants Copilot to Run Your PC—Now It Has to Earn Your Trust
Microsoft’s latest AI-PC push arrives with baggage. Its earlier Copilot+ PC campaign stumbled when Recall, the headline feature, was delayed; that history hangs over a new effort to make Windows machines both more capable locally and more willing to act on a user’s behalf.
At its San Francisco Windows and Surface event, Microsoft unveiled the Surface Laptop Ultra, an Nvidia RTX Spark-powered machine starting at $2,599 and due October 16. The company also opened preorders for the $5,999 Surface RTX Spark Dev Box, aimed at developers who want to run local AI models—including models exceeding 120 billion parameters—rather than send every demanding workload to the cloud.
The hardware is the foundation for what Microsoft calls “Hybrid Intelligence”: a mix of local and cloud models. Satya Nadella framed the shift as a new Windows chapter, saying PCs would become places where agents can work “securely on your behalf.” Microsoft’s argument is straightforward: local processing can make AI more responsive, more private in some cases, and cheaper for developers whose appetite for model use exceeds cloud budgets.
But the consequential change is in Windows itself. Over the next couple of months, Copilot is set to gain permission-based access to local files and the ability to take actions across the operating system. In Microsoft’s onstage tax-filing demonstration, Autopilot searched folders for documents, renamed and zipped them, then drafted an email to an accountant with the attachment ready to send.
Microsoft says users retain oversight. Its new Microsoft Execution Containers framework is intended to confine agents to sandboxes, give administrators visibility and set limits on what each can access. Yet that safeguard underscores the anxiety it is meant to answer: handing an AI bot the keys to files, email and system controls is a markedly bigger leap than asking it to summarize text. As one account of the event put it, “questions about security and user acceptance still remain,” especially after past mistakes and wider agentic-AI mishaps.
Microsoft is betting that local AI will make agents useful enough to justify that leap. Whether Windows users agree may matter more than how many parameters its new machines can run.
Write a comment