According to the researchers, once the extension was installed, it could reach each product's built-in AI with just a single click. The affected products span multiple vendors, which makes the finding notable: the same basic technique worked across all five environments rather than being limited to one vendor's implementation.

The report describes the affected products as:

• Google Chrome (Gemini Live)
• Perplexity Comet
• Microsoft Edge
• Opera Neon
• Claude in Chrome extension

The research highlights a broader concern for anyone running AI assistants inside the browser. Because these assistants often operate with access to page content and user context, an extension that can hijack them inherits that same reach. Users who have installed extensions they do not fully trust, or who granted broad permissions during installation, are the most exposed.

At this stage, the article does not name a specific CVE or vendor advisory for the issue, so no identifier is cited here. Readers should watch for vendor responses and updated extension-permission guidance from each of the affected products.

Source: The Hacker News

Do you run any of these AI assistants in your browser, and if so, are you reviewing your installed extensions in response?