Chrome Gemini Vulnerability
A newly discovered high-severity vulnerability in Google Chrome’s Gemini Live integration, tracked as CVE-2026-0628, exposed users to significant privacy and security risks.
Researchers found that the flaw could allow malicious browser extensions to hijack the Gemini side panel, granting unauthorized access to a user’s camera, microphone, and local files.
The integration of AI assistants into web browsers, creating so-called “agentic browsers,” fundamentally changes the browser security landscape.
Features like Gemini Live in Chrome require deep, privileged access to the browsing environment to perform tasks like real-time content summarization and automated execution.
| CVE ID | Severity | Affected Component | Exploit Mechanism | Impact | Status |
|---|---|---|---|---|---|
| CVE-2026-0628 | High | Google Chrome Gemini Live Panel | JavaScript injection via declarativeNetRequests API | Unauthorized access to camera, mic, files, and screenshots | Patched (Jan 2026) |
This multimodal perspective, where the AI sees what the user sees, creates a widened attack surface.
The privileged nature of these AI components means that vulnerabilities within them can bypass traditional browser security models.
Researchers at Palo Alto Networks Unit 42 found that the core issue with CVE-2026-0628 lies in how Chrome handles the Gemini web app (hxxps[:]//gemini.google[.]com/app) when loaded within the new side panel versus a standard tab.
By design, browser extensions using the declarativeNetRequests API can intercept and modify HTTPS requests, a feature often used by ad blockers.
While injecting JavaScript into the Gemini app in a regular tab doesn’t grant special privileges, doing so when the app is loaded in the Gemini panel is highly dangerous.
Chrome grants the panel elevated capabilities, such as reading local files and accessing media devices, to enable complex AI tasks.
By intercepting the app within the panel, an attacker could hijack these elevated privileges.
Exploiting this vulnerability would allow an attacker to execute arbitrary code within the privileged context of the Gemini panel. This could lead to severe consequences, including:
These actions could be performed with minimal user interaction, requiring only that the user open the Gemini panel.
Unit 42 responsibly disclosed the vulnerability to Google in October 2025, and a patch was released in early January 2026.
This incident highlights the novel security challenges introduced by AI browser integrations and the importance of continuous monitoring as these technologies evolve.
Follow us on Google News, LinkedIn, and X to Get Instant Updates and Set GBH as a Preferred Source in Google.
Attackers abused compromised ConnectWise Cloud Backup credentials to access Microsoft 365 customer environments, prompting ConnectWise…
Anthropic says an internal Claude research model reconstructed how the National Security Agency generated seeds…
Two WordPress Core vulnerabilities can be chained to execute remote code without authentication, according to…
A newly disclosed OpenBSD kernel exploitation chain combines a shared memory accounting flaw with a…
GhostAction attackers have expanded their GitHub credential theft campaign with a second stage that disguises…
Indian authorities have warned that SideCopy hackers are targeting Indian organizations, including academic institutions, with…