Microsoft on Tuesday unveiled a new AI-based web content analysis tool named Copilot Vision, underscoring safety and security to address potential concerns.

The tech giant seems to have learned its lesson from the launch of the Windows Recall feature, for which it pulled previews in June over security and privacy concerns. 

Last week, Microsoft announced the return of Recall after adding proof-of-presence encryption, anti-tampering checks, and secure enclave data management.

On Tuesday, the company unveiled several new and enhanced features for its Copilot generative AI offering. The list includes Copilot Voice for easier interactions with the chatbot, Copilot Daily for a morning summary of news and weather, Copilot Labs for testing experimental features, and Copilot Vision.

Copilot Vision, which Microsoft is currently testing in Copilot Labs, is a feature built into the Edge browser that sees what the user sees while they are surfing the web. Users can have conversations with the AI chatbot about what is on the screen and ask questions about the displayed content, including text and images.

“If you want it to, [Copilot Vision] can understand the page you’re viewing and answer questions about its content. It can suggest next steps, answer questions, help navigate whatever it is you want to do and assist with tasks. All the while you simply speak to it in natural language,” Microsoft explained. 

In an attempt to avoid raising security and safety concerns as in the case of the Recall feature, the tech giant has highlighted several aspects related to Copilot Vision. 

The feature is entirely opt-in and each session is ephemeral. None of the content the assistant engages with is stored or used for training — it’s permanently deleted as soon as the session ends. 

In addition, there are boundaries on the types of sites Copilot Vision can engage with, and at first it will only work with a limited list of popular websites. Microsoft has also highlighted that Copilot Vision will not work on paywalled and sensitive content.

“There is no specific processing of the content of a website you are browsing, nor any AI training. Copilot Vision simply reads and interprets the images and text it sees on the page for the first time along with you,” the company explained. 

Microsoft also added, “Before we launch broadly, we’ll continue to take feedback on all the above from early users in Copilot Labs, refine our safety measures, and keep privacy and responsibility at the center of everything we do.”

Related: Worldcoin: Fighting Deepfakes and Bots With Global Permissionless Blockchain Identity

Related: California Governor Vetoes Bill to Create First-in-Nation AI Safety Measures

Related: Critical Nvidia Security Flaw Exposes Cloud AI Systems to Host Takeover