Microsoft’s latest Copilot feature brings AI vision into Edge, letting it "see" the page you’re on and assist with questions or tasks as you go.
Key Points:
- Vision can summarize text, answer questions, and highlight products on select sites
- Requires a Copilot Pro subscription and limited to US for now
- Microsoft says it deletes all data after each session
Microsoft is rolling out Copilot Vision, a feature it calls “the first AI experience of its kind, available only on Microsoft Edge,” in a careful, invite-only preview for some Pro-level subscribers in the United States. The company’s AI assistant can now see what’s on your screen and provide insights in real time. Think of it as a companion that can decipher a complicated webpage, point out product deals, help you navigate a lengthy recipe, or translate a tricky paragraph—all while you browse.
This new capability doesn’t automatically kick in. Users must opt in, making Vision an entirely voluntary tool. Microsoft emphasizes that, once you end a session, it deletes the data associated with that session. It also promises that Copilot Vision doesn’t tap into your browsing data to train its AI models. The initial rollout includes only a small list of pre-approved websites, ensuring the tool’s reach is limited and controlled.
The company acknowledges privacy and copyright concerns. Publishers and news outlets have worried about bots scraping content and bypassing paywalls. Copilot Vision will not operate on paywalled or “sensitive” pages and will respect “machine-readable controls on AI,” according to Microsoft. The definition of “sensitive” remains vague, but Microsoft says it aims to expand the list of supported sites over time, taking feedback from publishers and users.
It’s a delicate balance. Microsoft wants to offer users a helpful AI browsing companion without running afoul of content owners and privacy laws. Copilot Vision’s first tentative steps are meant to reassure skeptics, refine the technology, and gradually gain trust. The ultimate goal is to make browsing more interactive and less isolated, turning a static webpage into a shared problem-solving space between user and AI.