How to Spot a Malicious AI Browser Extension Before It Steals Your Data
Two Chrome extensions impersonating a popular AI assistant reached a combined 900,000 installs, then quietly scraped complete ChatGPT and DeepSeek conversation histories to an attacker's server every 30 minutes, all while telling users in their store listings that they only collected "anonymous, non-identifiable analytics data." That's not an isolated incident. Here's what's actually happening with fake AI extensions right now, and how to check what's already sitting in your browser.
Browser extensions run with deep, persistent access to what you do online by design, not by exploiting a flaw — which is exactly why a malicious one is so effective.
Check Permissions, Not Just Reviews
The single most useful habit is comparing what an extension actually does against what it asks to access. If a "ChatGPT sidebar" extension requests permission to read and modify data on every site you visit, that's broader access than a sidebar needs, and it's exactly the kind of permission scope these campaigns rely on. Star ratings and install counts don't protect you here — several of the malicious extensions Microsoft and independent researchers found had hundreds of thousands of installs and functioned normally on the surface the entire time.
Go to your browser's extensions page right now (chrome://extensions in Chrome) and look at what's actually installed. Chrome's own Safety Check feature will flag extensions suspected of malware or removed from the store, and it takes under a minute to run.
Why AI-Branded Extensions Became a Favorite Target
Reviewing installed extensions and their permissions takes a few minutes and is the single most effective check available to an individual user.
Malicious updates can also arrive after installation — extensions update silently, with no new consent prompt, which is a separate risk from the initial install decision.
The pattern is straightforward: attackers build or acquire an extension, brand it as a sidebar or companion for a popular AI tool like ChatGPT, Claude, or DeepSeek, and list it in the Chrome Web Store or a similar marketplace. Because these categories are heavily searched right now, impersonation extensions get discovered and installed quickly, sometimes accumulating hundreds of thousands of users before anyone catches the behavior underneath.
One documented case used DOM scraping, reading the actual content rendered on a page, to pull complete AI chat conversations and browsing history, then sent that data to an attacker-controlled domain at a fixed interval. Separately, researchers disclosed a technique called HashJack, where adversarial instructions hidden in a URL's hash fragment can cause an AI browser assistant to leak data or expose credentials without exploiting any flaw in the extension itself, just by visiting a page with the malicious fragment.
A related risk applies even to extensions that were legitimate when you installed them: browser extensions update automatically and silently, with no new consent dialog. A developer selling their extension, or a compromised developer account, is a documented channel through which a previously safe tool has turned malicious overnight, well after the original install decision was made in good faith.
Full-Page Access
Chrome's permission system can legally grant extensions access to tabs, page content, form inputs, and cookies — broad access is often normal, not automatically a red flag on its own.
Wildcard Host Permissions
A permission granted with a blanket pattern extends that access to every single site you visit, not just the ones the extension needs to function.
Silent Auto-Updates
New extension versions install in the background with no fresh consent screen, which is how a safe extension can become unsafe after the fact.
Ownership Changes
Extensions occasionally change hands to new developers, sometimes without users noticing, which can precede a shift toward malicious behavior.
How to Check What's Already In Your Browser
Open Your Extensions Page
In Chrome, go to chrome://extensions. In Firefox, about:addons. Look at everything listed, including ones you forgot you installed.
Use Safety Check
Chrome's Safety Check flags extensions suspected of malware, ones removed from the store, or those with unclear privacy practices — it's built in and takes seconds.
Read What Each One Actually Requests
Click "Details" on each extension and check its permissions against what it claims to do. A summarizer needing access to every site and your full browsing history is a mismatch.
Quick Search the Extension Name
Search the extension's exact name plus "malware" or "security" before installing anything new — security researchers and tech outlets publish warnings that don't always show up in the store listing itself.
What You Can Control and What You Can't
- What you install, and how carefully you check permissions before installing it
- Running periodic audits of your extension list, since old, unused extensions are pure risk with no benefit
- Keeping browser Safety Check features turned on and actually reading their output
- Removing anything you don't recognize or no longer use, immediately
- Whether a currently-safe extension's developer account gets compromised or sells to a bad actor later
- Automated store review catching every malicious extension before it reaches meaningful install numbers
- Silent auto-updates changing an extension's behavior after your original install decision
- Data broker demand for browsing and AI-chat data, which is the underlying reason this category keeps getting targeted
Is This Worth Actually Doing Right Now?
Yes — it's a five-minute check with a real, documented threat behind it.
This isn't a hypothetical risk being used to sell you something. Microsoft, independent security researchers, and multiple threat intelligence firms have all documented real campaigns with real install numbers targeting AI-branded extensions specifically over the past year. The fix costs almost nothing: open your extensions page, run Safety Check, remove anything unfamiliar or unused, and read permissions before installing something new. None of that requires technical expertise, just five minutes of attention.
Frequently Asked Questions
How do I know if a browser extension is stealing my data?
There's usually no visible sign — malicious extensions are designed to function normally on the surface while quietly exfiltrating data in the background. The most reliable check is reviewing each extension's actual permissions in your browser's extensions page and asking whether that access level makes sense for what the tool claims to do. An extension that only claims to summarize pages but asks for access to every site you visit and your browsing history is a mismatch worth investigating.
Are AI browser extensions specifically more risky than other extensions?
The category has drawn a disproportionate amount of malicious attention recently, largely because it's currently one of the most heavily searched and installed extension types, which makes it an efficient target for attackers impersonating popular tools. Microsoft documented malicious AI-themed extensions reaching roughly 900,000 installs across more than 20,000 enterprise accounts. The risk isn't inherent to AI extensions as a category, but the current volume of impersonation attempts specifically targeting AI assistant branding is real.
What should I do if I find a suspicious extension already installed?
Remove it immediately through your browser's extensions page, then change passwords for any sensitive accounts you accessed while the extension was active, particularly email, banking, and any AI chat platforms where you may have discussed sensitive information. If the extension had broad permissions and you used it for an extended period, it's worth reviewing recent account activity for anything unfamiliar.
More Software Guides, Written Clearly
Browse our full library of software and security explainers, updated as new threats and tools emerge.
More from the creator
Explore two other projects from the same creator.