Security researchers at Forever Security have shown that a single, seemingly ordinary browser extension could gain control of the AI assistants embedded in five Chromium-based products. The affected targets span Google Chrome's Gemini Live, Perplexity Comet, Microsoft Edge, Opera Neon, and the Claude extension for Chrome. The finding suggests that a common architectural weakness in Chromium-based browsers may allow a malicious extension to intercept or manipulate AI-driven responses across different vendors.
Because the research is described only in a brief report, the precise technical steps are not yet public. However, the implication is clear: users who install third-party extensions may unknowingly expose their AI interactions to hijacking, even if the extension itself appears benign. The researchers did not indicate whether any of the affected companies have issued patches, and no fixes were mentioned in the source.
This is a single-source report, so there are no independent confirmations or differing views to compare. Still, the breadth of affected products highlights a systemic risk: as AI assistants become standard features in browsers, the extension ecosystem—already a known attack surface—could become a gateway for compromising these new capabilities. Users should treat extension permissions with caution, especially those that request broad access to page content or browser APIs.