The browser extension that promised to speed up your workflow might be doing something else entirely. Over the past year, security researchers have documented a pattern that should make any remote worker think twice before clicking “Add to Chrome.” A single extension marketed as a privacy VPN accumulated more than six million installs across Chrome and Edge before investigators discovered it had been silently capturing every prompt and response people typed into AI tools like ChatGPT, Claude, and Gemini — even when the VPN was turned off. The data wasn’t just collected; it was sent to a data broker that sells behavioral profiles to advertisers. This isn’t a theoretical risk. It’s happening now, at scale, and the browser you use all day for work is the delivery mechanism.
Browser Security
Data Privacy
Remote Work Risk
AI Tools
The Extension You Trusted May Be Watching Your AI Conversations
Most of what’s documented here comes from a deep investigation by security researchers at Koi, reported by Cyberpress, and a separate threat advisory from UVCyber’s TIDE team. Both groups independently confirmed that the Urban VPN Proxy extension — which held a Google “Featured” badge and was promoted as a privacy tool — began quietly intercepting AI conversations starting with version 5.5.0, released July 9, 2025. The extension injected JavaScript into browser networking functions, overriding standard APIs like fetch() and XMLHttpRequest to capture every message before it even reached the AI platform’s interface. Prompts, responses, timestamps, and session metadata were then transmitted to servers at analytics.urban-vpn.com and stats.urban-vpn.com via background workers.
The harvesting continued regardless of whether the user had disabled the extension’s so-called “AI protection” feature. And it wasn’t just one extension: researchers found that 1ClickVPN Proxy, Urban Browser Guard, and Urban Ad Blocker contained the same embedded scripts. Combined, more than eight million users were affected. The data ended up with BiScience (B.I. Science Ltd.), an affiliated data broker that sells behavioral and browsing data through products AdClarity and Clickstream OS.
Urban VPN marketed its data collection as a security feature that “scans prompts for personal information or suspicious links.” Disabling it did nothing — the interception was hardcoded into the extension’s core functionality. This is a textbook example of a feature being designed to benefit the vendor, not the user.
How Extensions Get Dangerous After They’re Installed
The Urban VPN case isn’t an outlier. It’s part of a broader shift in how browser extensions become threats. Unlike traditional malware that exploits software flaws, malicious extensions operate within the permissions the user already granted. They update automatically, often without any visible change in behavior. And they can stay benign for years before a single update flips them into data-harvesting mode.
Security researchers at Barracuda describe these as “sleeper agents” — extensions that accumulate trust signals like high download counts, positive reviews, and longevity in official stores, then weaponize through a later update. The DarkSpectre campaign, for example, kept extensions clean for more than five years before activating malicious code, eventually infecting an estimated 8.8 million users. In another case documented by Bitdefender, researchers discovered 108 Chrome extensions tied to a single command-and-control server, disguising themselves as Telegram helpers, translation tools, and slot games. Those extensions stole Google account data, exfiltrated Telegram Web sessions, and opened arbitrary URLs at browser startup — all while appearing to function normally.
Audit Your Extension List
Open your browser’s extension manager and review every installed add-on. Remove anything you haven’t used in the past 30 days. Pay special attention to extensions that request broad access to “read and change all your data on websites.”
Check Permissions Against Functionality
An ad blocker doesn’t need access to your browser’s networking APIs. A VPN extension shouldn’t be intercepting your AI chat sessions. If a tool’s permissions seem disproportionate to what it does, that’s a red flag.
Treat Updates as New Decisions
When an extension updates, its behavior can change entirely. If you see a notification that an extension has been updated — especially one you’ve had for a while — consider removing it and finding a newer, actively maintained alternative.
Why This Matters Specifically for Remote Workers
Remote work concentrates almost everything in the browser. You authenticate into SaaS platforms, collaborate on documents, manage projects, and increasingly use AI to draft communications, review code, summarize research, and test strategic ideas. Each of those activities leaves a trail inside the browser, and any extension with the right permissions can read it. The HCLTech blog notes that a few weeks of AI conversations can reveal more about a company’s priorities and direction than months of generic browsing history. For a remote worker, that means proprietary code, confidential customer details, internal strategy notes, and even legal analysis can be quietly exfiltrated without any obvious performance hit or alert.
What makes this especially tricky for remote workers is the absence of corporate IT oversight. On a company-managed device, an administrator might restrict extension installation to an approved list. But many remote workers use personal devices for work tasks, or they install extensions on a work laptop without thinking twice. The browser becomes a blind spot that grows in proportion to AI adoption.
What You Can Do Right Now (Without Becoming a Security Expert)
You don’t need to become a cybersecurity specialist to reduce your exposure. A few straightforward habits can close the most common gaps.
- Keep your work browser separate from your personal browser. Use one for company tasks and a different one (or a separate profile) for everything else. This limits any compromised extension’s access to sensitive data.
- Uninstall any Urban-branded extensions immediately. Researchers at Cyberpress and UVCyber both advise assuming that all AI conversations since July 2025 have been exfiltrated if you had one installed.
- Treat AI prompts as sensitive business content. Don’t paste proprietary code, customer PII, or internal strategy into a public AI interface unless you’ve verified that no extension has access to the page. A simple rule: if you wouldn’t email it to a stranger, don’t type it into an AI chat while an unknown extension is running.
- Review your extension permissions quarterly. A regular privacy check can catch extensions that have changed their behavior or that you no longer need. This quick checklist covers the basics for a home office setup.
- Consider using endpoint security software that monitors browser extension behavior. Some antivirus suites now include browser-specific protections that flag unusual outbound data flows from extension processes. (Just make sure the security tool itself doesn’t come with its own questionable extension.)
The Bigger Picture: Trust That Must Be Earned Daily
The browser extension problem isn’t a technical glitch that can be patched away. It’s a structural consequence of how we’ve built the modern web — where a few lines of JavaScript installed with a single click can sit between you and every website you visit. The same convenience that makes extensions so useful also makes them a perfect channel for data harvesting. And because extensions update automatically and silently, the trust you placed in a tool six months ago may already be invalid.
That doesn’t mean you should uninstall every extension and work in a bare browser. It does mean treating extensions as what they are: third-party software running inside your primary workspace. The same scrutiny you’d give to downloading an app on your phone should apply to a browser extension, especially one that claims to improve productivity or privacy. If a tool is free and its business model isn’t obvious, the product might be your data.
What’s encouraging is that the defenses are straightforward. You don’t need a dedicated security team. You need a habit of looking at what’s installed, questioning permissions, and periodically cleaning out the clutter. The browser is the front door of remote work. Keeping it secure is less about fear and more about paying attention — the same way you’d lock your home office door when you step away.