What data leaves our machines?

One JSON summary per endpoint: browser, profile, username, extension ID and version. The collector reads extension folders, the preferences file that says whether an extension is enabled, and Firefox and Safari's equivalents. Never page content, browsing history or credentials, and it writes nothing to the endpoint. Stored in the EU, encrypted at rest. Disconnect and we stop syncing and forget the token.

What gets pushed into my SentinelOne console?

One indicator per extension on your fleet that we have confirmed malicious or rated high or critical, keyed by the extension ID so re-pushes update in place rather than duplicating. Medium findings stay out, and so do extensions you already block. No inventory yet means no indicators, not a dump of our public catalogue. We deliberately do not push per-domain network indicators, which are noisy and false-positive prone.

What happens on my endpoints when an indicator matches, and how do I remove one?

Nothing runs and nothing gets killed. The indicator is a URL pointing at the extension's store listing, so a match is a detection event, not a mitigation. We never push hashes or process indicators, so there is nothing for the agent to act on. Everything we write carries the source Am I Being Pwned, so you can filter or bulk-purge our entire contribution without touching your other feeds. Disconnect and our indicators stop refreshing and age out inside 90 days - we do not delete them from your store for you.

Can Am I Being Pwned remove or disable an extension on my endpoints?

No. Nothing here can touch a machine: the role has no endpoint action permissions, the collector only reads, and the push writes to your indicator store rather than to a policy. We produce the verdict and the evidence; you remove things with whatever manages your browsers already. Connect Chrome Browser Cloud Management separately and an admin can push a block from our dashboard - always an explicit human action, never an automatic sweep.

Does this work alongside Google Workspace or our MDM?

Yes. SentinelOne endpoints, Chrome Browser Cloud Management, our own extension and anything else merge into one view, deduplicated by extension ID: three sources seeing one extension is one row with the device counts attached, not three entries. Device counts stay split by source, because a machine your EDR and your browser management both see is one machine with two names - we show you both rather than guessing.

Integration

Get real visibility over the extensions installed across your fleet.

SentinelOne covers the endpoint. It does not cover browser extensions - trusted code inside a signed browser, reading every page your staff open, with no process to kill and no file to quarantine.

AIBP closes that gap across the same fleet your agents already manage: a read-only Service User for the endpoint list, a read-only collector for the extension census, and every malicious extension we find written straight back into your Threat Intelligence store.

You keep SentinelOne as the endpoint control plane. We become the browser layer feeding it.

Book a call Talk to an engineer

Connecting deploys nothing - read-only, account-scoped, revocable in one click

Acme fleet - 1,486 devices

Screenshot & AnnotateCRITICAL

Publisher account transferred to a new owner two months ago, with no changelog since.

The next release added a remote-config fetch that eval()s the JavaScript it gets back.

Beaconing observed to a domain registered 19 days before that update shipped.

312

devices

288

users

6

S1 sites

The gap

Your endpoint layer is owned. Your browser layer might not be.

Extensions provide value and risk

What SentinelOne already stops

Malicious binaries and scripts

Ransomware and lateral movement

Anything that runs as a process

What a browser extension does instead

Runs inside a signed browser

Reads pages and session tokens

Updates itself silently

Appears in no application inventory

The analysis

A permission list is not a verdict.

“Can read all sites” describes almost every useful extension. So we do not score permissions - we take the exact bundle running on your fleet apart, then run it and watch what it actually does.

Staticthe code as shipped

We analyse the exact version your fleet is running, not whatever the store ships today.

Packed bundles normalised before anything is judged
Declared permissions checked against the APIs actually called
Page content and cookies traced through to network sinks

Dynamicthe code as it behaves

Static analysis cannot see what a loader fetches at runtime, so we install the extension in an instrumented browser and drive it.

An agent drives it like a person: webmail, admin console, banking
Every DOM read, cookie access and outbound request recorded
Region, delay and domain varied to surface conditional payloads

Proof

Findings we published

MultiPassword

A password manager with over a million users leaking usernames, passwords and TOTP codes. CVSS 8.3. The manifest was clean.

Read the breakdown

Urban VPN

Taking commands from any website through an unvalidated postMessage handler.

Read the breakdown

Stylish

Two million users, exfiltrating every URL visited and the contents of AI chats.

Read the breakdown

Our founder is on the Centre for Cybersecurity Belgium's Wall of Fame for coordinated disclosure. Read more.

What changes for you

Four things you can do on Monday that you cannot do today.

Answer any extension question in seconds

“Is this on our fleet? Who has it? Since when?” - resolved from a search box instead of a three-day thread with IT and a half-trusted spreadsheet.

Search by extension name, ID or device / Device list attached / CSV export

Hand the auditor an artefact, not a dashboard

One click freezes the fleet: every extension, its verdict, the policy in force. Timestamped, signed off by name, and never rewritten - the evidence you currently write “N/A” against.

Frozen snapshot / CSV and print / Says who ran it, and when

Feed Singularity what it cannot see

Confirmed malicious, high and critical extensions land in your Threat Intelligence store hourly, each linked back to the full breakdown. Medium stays out: that is a policy review, not an indicator.

One indicator per extension / Deduped by extension ID / 90-day rolling expiry

Catch the silent update, not just the install

Same name, same icon, new bundle. We diff every sync and re-analyse the change - including the extensions that fetch their real instructions from a server after install.

Version diff each sync / Remote-config flagged / Re-analysed on change

Setup

Two ways to integrate.

The same read-only collector either way, so the same analysis and the same findings. Most teams deploy it themselves through tooling they already run; if you would rather not, we drive it through RemoteOps.

Manual collector

Default

A read-only PowerShell and bash collector with a write-only ingest key baked in. Plain text, short enough to read before you ship it, writes nothing to the endpoint. Deploy it like any other script: RemoteOps, Intune, Jamf, GPO or any RMM.

Fits your existing pipeline - the collector is just another script your tooling already ships

Narrowest credential - the token we hold reads your agent list and writes indicators

You own the rollout - ring it like any script, on the cadence your change process wants

Automatic via RemoteOps

Hands-off

If you would rather not own the deployment, we upload the same read-only collector to your RemoteOps script library and run it fleet-wide once a day, collecting results as agents come online. Requires the Singularity RemoteOps add-on.

Zero deployment work - nothing for your endpoint team to package or schedule

Offline endpoints included - results are collected as agents reconnect

Costs you a wider credential - upload and run means our token could execute anything on your fleet. Got a script pipeline already? Take the manual one

FAQ

Questions we get asked first.

No. SentinelOne's application inventory reads the Windows registry, macOS Spotlight, and Linux package managers, and browser extensions are not registered in any of them. There is no dedicated extension inventory or assessment page in the Singularity console, so extension enumeration today is a manual Deep Visibility query that returns IDs without a verdict.

Reviewing us as a vendor? Security and privacy.

Find out what is actually installed.

Connecting takes twenty minutes and deploys nothing. Inventory and indicators land on the next hourly sync.

Book a call Talk to an engineer

SentinelOne, Singularity and the SentinelOne logo are trademarks of SentinelOne, Inc. Used for identification only; this integration is built by Am I Being Pwned.

Get real visibility over the extensions installed across your fleet.

Your endpoint layer is owned. Your browser layer might not be.

What SentinelOne already stops

What a browser extension does instead

A permission list is not a verdict.

Findings we published

MultiPassword

Urban VPN

Stylish

Four things you can do on Monday that you cannot do today.

Answer any extension question in seconds

Hand the auditor an artefact, not a dashboard

Feed Singularity what it cannot see

Catch the silent update, not just the install

Two ways to integrate.

Manual collector

Automatic via RemoteOps

Questions we get asked first.

Does SentinelOne already inventory browser extensions?

What data leaves our machines?

What gets pushed into my SentinelOne console?

What happens on my endpoints when an indicator matches, and how do I remove one?

Can Am I Being Pwned remove or disable an extension on my endpoints?

Does this work alongside Google Workspace or our MDM?

Find out what is actually installed.