Users of AI platforms like ChatGPT, Claude, Copilot, Gemini, and DeepSeek are unknowingly exposing their personal and sensitive information. This vulnerability is being exploited by malicious browser extensions that are secretly collecting and transmitting data to unidentified servers.
Rising Threat of Malicious Extensions
The proliferation of AI-related browser extensions, which have reached approximately 115 million users globally as of March 2026, presents an attractive opportunity for cybercriminals. These extensions, posing as helpful tools, are covertly gathering user data, as revealed in a report by analysts at G Data and shared with Cyber Security News (CSN).
The report identifies three specific extensions: Urban VPN, Smart Sidebar, and AI Assistant (now Chat AI). These extensions, despite having high user ratings on the Chrome Web Store, are engaged in unauthorized data collection activities.
Data at Risk
These malicious extensions pose a significant threat due to the nature of the data being intercepted. Users often share personal information, confidential business data, and even medical records with AI platforms. This information, once intercepted, can be used for nefarious purposes such as fraud, blackmail, or corporate espionage.
The extensions operate by injecting scripts into the browser, intercepting network requests, and extracting conversation data without disrupting the AI platforms’ functionality. This makes detection by users extremely challenging.
Specific Extensions and Their Methods
Urban VPN, one of the most notorious extensions, was found to include a script that harvested data from multiple AI platforms, even when the VPN was inactive. Similarly, Smart Sidebar used a script to monitor and capture interactions on platforms like ChatGPT and DeepSeek, sending the data to suspicious domains.
The third extension, AI Assistant, incorporated a hidden iframe to intercept user interactions, forwarding data to unverified external URLs. Despite its ‘Featured’ status on the Chrome Web Store, it employed deceptive methods to gather information.
To protect against such threats, G Data advises users to only install extensions from trusted sources and apply the Principle of Least Privilege, ensuring extensions have minimal access permissions. Regular audits of installed extensions and organizational restrictions on browser access to sensitive platforms are also recommended.
Indicators of Compromise
The report highlights several indicators of compromise, including specific malicious extension hashes and detection names, which can aid in identifying and removing these harmful extensions.
For ongoing updates on cybersecurity threats, follow Cyber Security News on Google News, LinkedIn, and other platforms.
