Microsoft has launched a new feature for its AI-powered Copilot in preview – Copilot Vision. It allows the AI companion to process visual information on web pages and engage in conversations about the content with the user.

The feature is currently rolling out to only a limited number of Copilot Pro subscribers in the United States through Copilot Labs. A Copilot Pro subscription costs $20.00 per month.

Vision is one of the first features to debut in Copilot Labs, Microsoft's testing ground for experimental AI projects, along with Think Deepeer. Microsoft initially announced Copilot Vision in October as they rolled out other features to the AI companion, like a redesigned interface, Voice, Copilot Daily, and Personalized Discover, to name a few.

The company says that the feature will initially be compatible with only a limited selection of websites as they're treading lightly. They have plans to expand support based on user feedback and usage trends.

Copilot Vision is currently only available in the Microsoft Edge browser, a trend Microsoft tends to follow when debuting products the other companies need to play catch up on; it did the same when Copilot was initially launched (back when it was known as Bing AI).

Copilot now sits neatly at the bottom of the Edge browser from where you can access Copilot Vision; the icon resembles an RSS feed. When enabled, it processes visual information on the page, allowing users to discuss the content with the AI. This addresses the challenge of users having to explain what they are seeing to Copilot, transforming web browsing from a "lonely experience," according to Microsoft.

Privacy is a key consideration with Copilot Vision. The feature is opt-in, giving users control over its activation. Once you enable Vision, it stays on throughout the browsing session until you turn it off manually.

All conversation data, including context shared with Copilot, is deleted at the end of each session. However, Copilot’s responses are logged to improve safety systems. Importantly, Microsoft states that Copilot Vision does not store or use publisher content to train its models, and it honors website controls like robots.txt.

This new feature builds on Microsoft's ongoing efforts to integrate AI into its products and services. While a previous attempt, the "Recall" feature, faced privacy concerns due to its screen-capturing functionality, Microsoft has taken a more privacy-conscious approach with Copilot Vision. Instead of capturing screenshots, the AI directly analyzes the webpage content, addressing some of the earlier criticisms. Moreover, it can only see your screen when you want it to, unlike Recall.

"It sees the page you're on, it reads along with you, and you can talk through the problem you're facing together.”— Microsoft

In a few demos shared by the tech giant, Copilot Vision is seen as helping a user plan a day of fun activities for their niece or another to choose a Christmas sweater. It can help you find a product that matches your requirements from the endless list of products and even help answer any questions about it. The experience reminds me of being on a video call with my friends as they help me choose something over screen share. But Copilot can analyze everything in an instant. It can even help simplify learning or play games.

The introduction of Copilot Vision reflects the growing trend of incorporating AI into web browsing, transforming how users interact with online content. The company will monitor usage data and public response during the preview period before making the feature widely available.