AISEC > Security News > Ai Security > Chrome Gemini Live Vulnerability: Extension Hijacking Risk 2026
March 02, 2026 #ai-security Source: SecurityWeek 3 min read · 602 words

Vulnerability Allowed Hijacking Chrome’s Gemini Live AI Assistant

Вразливість дозволила перехопити AI-асистент Gemini Live у Chrome

Context: Why This Matters Right Now

Google's Gemini Live—the real-time conversational AI feature built into Chrome—was supposed to be a controlled, secure experience. Instead, researchers discovered it could be compromised by malicious browser extensions with access to spy on conversations and exfiltrate user files. This isn't theoretical. It's a concrete vulnerability affecting millions of Chrome users who trusted this feature to handle sensitive queries.

And frankly, this exposes a deeper problem: AI assistants integrated directly into browsers inherit all the security baggage of the extension ecosystem.

What We Know

According to SecurityWeek, the vulnerability allowed threat actors to hijack Gemini Live's functionality through crafted Chrome extensions. The attack vector is simple: once a user installs a malicious extension—whether through social engineering, typosquatting, or legitimate-looking ads—the attacker gains the ability to intercept Gemini Live sessions.

The timeline matters here. This wasn't a fresh-from-the-lab discovery. By the time SecurityWeek reported it on March 2, 2026, the vulnerability had likely existed in the wild for some period, which means exposure windows are always a concern with chrome vulnerability cve assignments and their tracking.

No specific CVE identifier was immediately disclosed at publication, though Chrome's vulnerability rewards program typically catalogs these incidents within weeks. If you're tracking chrome vulnerability list updates or monitoring chrome vulnerability today alerts, this one should be on your radar.

How It Works

Here's the technical core: Gemini Live extensions have elevated permissions that allow them to access DOM elements, read user input, and communicate with Google's backend. A malicious extension can hook into these same permission levels, essentially sitting in the middle of the conversation.

Think of it like this.

User speaks to Gemini Live. Extension intercepts. Attacker captures audio or text transcripts. Files sent to Gemini? They're now accessible to the extension with the same privileges.

The attacker doesn't need to break encryption or exploit a flaw in Gemini's code itself. They're exploiting Chrome's extension permission model, which was never designed with AI assistants in mind. That's a design problem, not just a bug.

Why It Matters

Gemini Live users often ask sensitive questions: legal advice, medical concerns, financial planning, proprietary business strategies. These conversations, if intercepted, become intelligence gold for competitors, blackmailers, or state-sponsored actors. And that's before we get to the file theft angle.

The real question is: how many users installed extensions without understanding the permission warnings? Chrome's extension permission UI is notoriously vague, and users routinely grant broad access to avoid friction.

This is particularly nasty because victims wouldn't necessarily know they'd been compromised. A malicious extension can operate silently in the background, logging conversations indefinitely. Unlike a phishing attack or ransomware incident, there's no obvious indicator of breach until forensics uncover the damage.

Next Steps

First, audit your installed extensions immediately. Remove anything you don't actively use or recognize. Don't trust extension names alone—check the publisher, review count, and last update date.

Second, if you've used Gemini Live extensively while untrusted extensions were installed, assume your conversations may have been captured. Review what you discussed and take protective steps: notify relevant parties, change passwords for services you discussed, and monitor accounts for suspicious activity.

Third, stay current on chrome vulnerability 2026 patches and chrome vulnerability update notifications from Google. Chrome's auto-update is your baseline defense, but it's not enough. Threat intelligence feeds like those tracking chrome vulnerability reddit discussions or official chrome cyber security advisories from Google will surface new risks faster than waiting for your browser to prompt you.

Finally, consider whether Gemini Live's convenience is worth the attack surface it introduces. For sensitive conversations, using the web interface in a sandboxed profile or separate browser instance limits extension exposure. It's not perfect, but it's better than letting every installed extension into the room.

Read original article →

// FAQ

Can malicious Chrome extensions spy on my Gemini Live conversations?

Yes. The vulnerability allows extensions with sufficient permissions to intercept Gemini Live sessions, capturing conversation transcripts and accessing files users share with the AI. Removing suspicious extensions and auditing your installed extensions immediately is recommended.

What's the CVE number for the Chrome Gemini Live hijacking vulnerability?

No specific CVE was publicly disclosed at the time of SecurityWeek's March 2, 2026 report. Google's chrome vulnerability rewards program typically catalogs these within weeks—check official Chrome security advisories for the assigned CVE.

How do I know if a malicious extension hijacked my Gemini Live?

There's no obvious indicator to users. Silent interception leaves no alerts or logs visible to the victim. Review your installed extensions for unfamiliar or unused ones, check browser activity logs, and consider credential rotation for sensitive accounts discussed with Gemini if you can't rule out compromise.

// RELATED

#ai-security Hackers Weaponize Claude Code in Mexican Government Cyberattack 2026-03-01 #ai-security ClawJacked Flaw Lets Malicious Sites Hijack Local OpenClaw AI Agents via WebSocket 2026-02-28 #ai-security ScarCruft Uses Zoho WorkDrive and USB Malware to Breach Air-Gapped Networks 2026-02-27

Concerned about your project's security? Run an automated pentest with AISEC — fully automated AI-powered scanner. Go to dashboard →