Copilot Vision, Microsoft's AI tool that can read screen content, launches preview version

In early October, Microsoft showed off CopilotVision, a new feature that harnesses the power of artificial intelligence to help you complete a variety of tasks by analyzing the content on your screen and providing necessary assistance. Two months after its initial release, Microsoft has finally launched a preview version of CopilotVision.

When enabled, CopilotVision will "see" and understand what is happening on the screen and its context. You can ask Copilot questions about web pages, and the assistant will answer and provide more information.

CopilotVision is currently only offering a limited preview to CopilotPro users in the United States. Additionally, the feature only works on Microsoft Edge browser and only on certain websites. Microsoft says it wants to make sure Vision is a "useful and secure experience" from the start, but the company plans to expand the number of available websites as testing expands.

Microsoft also emphasized that CopilotVision does not collect or store any user data or use it for model training. After the session ends, CopilotVision will delete all data related to what you said or shared information. Like the recently launched Recall (also in preview for Windows Insiders using Copilot+PC), CopilotVision is an optional feature -- it won't be called unless you enable it.

In addition, Microsoft is committed to putting copyright, privacy and security first. It does not provide any information about paid sites and cannot take action on your behalf. CopilotVision's sole purpose is to "have a second pair of eyes", it's a helpful assistant that tells you more information no matter what you ask.