Malicious VPN Extension Stole ChatGPT and Google Gemini Conversations From Millions

A malicious VPN service was found stealing full ChatGPT and Google Gemini conversations, raising major concerns about browser extension privacy and AI data security.

Popular VPN extension caught harvesting AI conversations

A recent investigation revealed that the VPN extension Urban VPN Proxy secretly logged full browser traffic, including private conversations with major AI platforms. The affected services include ChatGPT, Claude, Google Gemini, Microsoft Copilot, Perplexity, DeepSeek, Grok (xAI), and Meta AI.

Despite carrying a 4.7-star rating from 58.5K reviews and reaching over six million users, the extension exposed sensitive data from more than eight million users worldwide. Google even featured the extension, which increased user trust and adoption.

How Urban VPN stole ChatGPT and Gemini data

Urban VPN collected data by injecting an executor script directly into the webpages of targeted AI platforms. The script overrode native browser functions, which allowed it to intercept all network traffic between users and AI services.

The extension captured:

• Every AI prompt and response
• Conversation IDs and timestamps
• Session-level metadata

After extraction, the script compressed the data and transmitted it to Urban VPN servers. Data harvesting ran continuously in the background and worked even when the VPN was turned off or user settings were unchanged.

All conversations since July 2025 are compromised

The malicious functionality appeared in Urban VPN version 5.5.0, released on July 9, 2025. Any AI conversations conducted since that date should be treated as compromised.

Urban VPN is affiliated with BiScience, a data broker company. Reports indicate that the harvested data was collected and sold for marketing analytics purposes.

Other extensions affected

Urban VPN was not the only extension involved. Investigators found the same data-harvesting code in:

• 1ClickVPN Proxy
• Urban Browser Guard
• Urban Ad Blocker

Users running any of these extensions should remove them immediately and review their online accounts for potential exposure.

Many users rely on AI chats for private tasks such as writing code, planning projects, or discussing sensitive topics. This incident shows how browser extensions can silently intercept AI conversations without clear consent.

The findings highlight the importance of auditing browser extensions regularly and avoiding VPN tools with unclear data-collection practices.

In other news, ChatGPT can now edit photos and PDF files, so if you removed the malicious extension, go give it a try.

Via Neowin