Skip to main content

Chrome Flaw Lets Malicious Extensions Spy Through Your AI Assistant

Chrome's AI Feature Becomes Security Weak Point

Security researchers have uncovered a disturbing flaw in Google Chrome that turns its helpful AI assistant into a potential spy. The vulnerability, tracked as CVE-2026-0628, gave malicious extensions a backdoor to sensitive system resources through the Gemini Live panel.

How the Hack Works

The attack begins when users install seemingly harmless extensions. These bad actors manipulate how Chrome handles sidebar requests, sneaking past normal security checks. Once they gain control of Gemini's interface - designed to assist users - they inherit its elevated privileges.

"It's like giving a burglar your house keys," explains Gal Weizman from Palo Alto Networks' Unit 42 team. "Because Gemini needs legitimate access to function normally, hackers can piggyback on those permissions."

The consequences are chilling: unauthorized camera activation, secret microphone recording, unrestricted file access, and even phishing attempts disguised as friendly AI conversations.

Patch Now Available

Google moved quickly to seal this security breach in January's update (version 143.0.7499.192). Users who haven't updated their browsers should do so immediately - this isn't a vulnerability you want lingering on your devices.

The incident has sparked debate among cybersecurity professionals about balancing convenience with protection. As browsers evolve into full-fledged operating systems with integrated AI capabilities, they're becoming juicier targets for attackers.

The Bigger Picture

Industry analysts had warned about these risks before. Research firm Gartner cautioned against "proxy" browsers that weave too deeply into system operations. While AI-powered features promise productivity boosts, they may create security holes that outweigh their benefits.

The Chrome incident serves as a wake-up call: every new technological advancement brings fresh vulnerabilities we must anticipate and guard against.

Key Points:

  • Critical Access Granted: Malicious extensions could bypass normal restrictions through Gemini Live panel hijacking
  • Multiple Privacy Risks: Successful attacks enabled camera/microphone activation and local file access
  • Patch Deployed: Update to Chrome version 143.0 or higher closes this security loophole
  • AI Integration Concerns: Deep system connections create new attack surfaces needing careful evaluation

Enjoyed this article?

Subscribe to our newsletter for the latest AI news, product reviews, and project recommendations delivered to your inbox weekly.

Weekly digestFree foreverUnsubscribe anytime

Related Articles

News

Claude AI Spots 100 Firefox Flaws in Record Time

In a cybersecurity breakthrough, Mozilla partnered with Anthropic's Claude AI to uncover over 100 Firefox vulnerabilities within two weeks. The AI detected 14 critical security risks along with numerous lesser issues, demonstrating superior efficiency compared to traditional testing methods. These findings have already been patched in Firefox's latest update.

March 9, 2026
CybersecurityAI InnovationBrowser Safety
Popular AI Tool OpenClaw Poses Serious Security Threat to macOS Users
News

Popular AI Tool OpenClaw Poses Serious Security Threat to macOS Users

The widely-used AI assistant OpenClaw has been found to contain critical vulnerabilities that hackers are actively exploiting. Security researchers warn that attackers are tricking macOS users into running malicious scripts disguised as legitimate setup instructions. Once executed, these scripts disable key security protections, allowing stealthy malware to steal sensitive data like passwords and developer credentials.

February 4, 2026
OpenClawmacOS SecurityAI Vulnerabilities
News

Nearly a Million Chrome Users Hit by Sneaky AI Chat Thieves

A startling security breach has exposed private conversations from ChatGPT and DeepSeek users through malicious Chrome extensions. Disguised as helpful tools, these add-ons secretly harvested sensitive data from over 900,000 unsuspecting victims. Security experts warn this sophisticated attack could compromise corporate secrets and personal information alike.

January 8, 2026
Chrome SecurityAI PrivacyData Breach
ChatGPT's ShadowLeak Vulnerability Exposed Gmail Data Theft
News

ChatGPT's ShadowLeak Vulnerability Exposed Gmail Data Theft

Security researchers discovered a critical vulnerability in ChatGPT's 'Deep Research' mode, dubbed 'ShadowLeak,' which allowed attackers to silently steal Gmail account data. The attack exploited OpenAI's cloud infrastructure, leaving no traces and bypassing local security measures. OpenAI has since patched the flaw, but the incident highlights ongoing risks in AI agent systems.

September 23, 2025
ChatGPTCybersecurityAI Vulnerabilities
Notion 3.0 AI Agent Vulnerability Exposes Sensitive Data via Malicious PDFs
News

Notion 3.0 AI Agent Vulnerability Exposes Sensitive Data via Malicious PDFs

A critical security flaw in Notion 3.0's AI agent feature allows attackers to exploit malicious PDFs to bypass protections and steal sensitive data. Cybersecurity firm CodeIntegrity warns that the vulnerability stems from weaknesses in LLM integration, tool permissions, and memory systems, affecting even advanced models like Claude Sonnet 4.0.

September 22, 2025
Notion SecurityAI VulnerabilitiesData Protection
Google Gemini Assistant Security Flaw Exposed
News

Google Gemini Assistant Security Flaw Exposed

A new study reveals critical vulnerabilities in Google's Gemini assistant, where attackers can exploit everyday emails and calendar invites to execute malicious commands. Researchers demonstrated risks like smart home control and data theft, prompting Google to implement security fixes.

August 8, 2025
CybersecurityAI VulnerabilitiesGoogle Gemini