Enterprise Browser Extensions Are the Attack Surface Nobody Is Managing

Most enterprise security programs can tell you exactly which laptops are enrolled in device management, which SaaS apps have single sign-on, and which USB ports are locked down. Ask the same security team how many browser extensions are installed across the company, and what each one can read, and the answer is usually a shrug. That gap is now one of the largest unmanaged attack surfaces in a typical office.

LayerX's 2026 Enterprise Browser Extension Security Report found that 99% of enterprise users have at least one extension installed, and roughly a quarter have more than ten. Extensions run with broad permissions by design — reading page content, intercepting network requests, injecting scripts — because that's what makes a password manager or a grammar checker useful. The same permissions make a compromised or malicious extension nearly as powerful as malware that's already inside your network.

A Concrete Example of How This Plays Out

In April 2026, researchers at Socket identified a coordinated cluster of 108 malicious Chrome extensions sharing command-and-control infrastructure, collectively installed roughly 20,000 times through the official Chrome Web Store. Of those, 54 harvested Google account identity data through abused OAuth flows, and 45 contained backdoor functionality. What made the cluster hard to catch wasn't some exotic exploit — it was the extensions using entirely legitimate Chrome capabilities: host permissions, content scripts, and the declarativeNetRequest API that's meant for ad blockers and privacy tools. Reviewing the code for "does this extension request broad host permissions" doesn't flag anything, because nearly every useful extension does.

Why the Scale of the Problem Is Disputed

Google's own telemetry says less than 1% of Chrome Web Store installs in 2024 involved malware — a number the company uses to argue its review process works. Academic researchers at Stanford and CISPA, studying the same ecosystem from the outside, estimated that 280 million people installed a malware-infected Chrome extension over a three-year period. Both figures can be true simultaneously: a small percentage of a store with tens of millions of daily installs and roughly 150,000 to 200,000 active extensions still adds up to a very large absolute number of exposed users. The size of the store itself is contested too — estimates range from about 112,000 to nearly 200,000 active extensions, which tells you how little systematic visibility even researchers have into what's actually being distributed.

AI Extensions Are a New, Faster-Growing Risk

The newest wrinkle is AI browser extensions — summarizers, writing assistants, and agent-style tools that read page content to feed into a model. LayerX's report found these extensions are 60% more likely to have a known CVE than the extension population overall. The reason is structural: an AI extension typically needs broader read access to page content than a traditional utility extension, and the rapid pace of AI tooling means these extensions ship and update faster than security review processes can keep up with. A browser extension that legitimately needs to read everything on every page in order to summarize it is, from a security standpoint, functionally identical to one designed to exfiltrate everything on every page.

What Actual Extension Governance Looks Like

The organizations handling this well aren't trying to ban extensions outright — that fails the moment someone needs a legitimate password manager. Instead, the practical control set looks like this: maintain an allowlist enforced through enterprise browser management (Chrome Enterprise, Edge for Business, or a dedicated enterprise browser), rather than relying on users to self-police; require that any extension requesting access to all sites go through a specific approval step rather than the default install flow; monitor for extensions that change their requested permissions after installation, since a legitimate extension sold to a new owner is a known vector for turning a trusted tool malicious; and treat AI extensions as a distinct review category with tighter default restrictions, given the CVE rate LayerX documented.

None of this requires exotic tooling — Chrome Enterprise's extension management policies and equivalent controls in Edge and Firefox's enterprise builds already support allowlisting and permission auditing. The gap is organizational, not technical: most IT teams built mature device and SaaS governance years ago and never extended the same discipline to the thing every employee uses dozens of times a day. Browser extensions got left out of the asset inventory, and the 108-extension cluster is a reminder of what sits in that blind spot.