AML-Filter

Free sanctions screening for small businesses: check a customer's name against government lists, in your browser.

Open the app at aml-filter.com. No sign-up, nothing to install.

If your business has to "know your customer", you have to check that the people you deal with are not on a sanctions list. Big companies pay a screening vendor for this. Small ones often search each government website by hand and keep notes in a spreadsheet. The vendor costs money and gets a copy of your customer list. The manual way is slow, and a name spelled a little differently is easy to miss.

AML-Filter is a free web page that does the checking for you. It searches the US, EU, UN and UK sanctions lists, looks for names that are spelled or sound similar, and shows why each result came up. It runs entirely in your browser. The names you type and your customer list stay on your computer.

Technical docs: Getting started for developers · Architecture · How matching works · Watchlist format · Deploy

Try it

1. Open aml-filter.com/screen. The first visit downloads the US sanctions list and a small name-matching model, about 70 MB. Later visits start fast.
2. Type Vladimir Putin and press Enter.
3. Click Why this score? under the top result.

The top result scores 0.952. Scores run from 0 to 1, and higher means a closer match. The breakdown shows what drove it: the names are close in meaning and in spelling, and "PUTIN, Vladimir" is one of the listed aliases. The results below it are weaker matches that share a first name. The tool never decides for you. A person looks at each result.

On a computer the search page checks all four lists, and each result names its list. On a phone it starts with the US list (OFAC) to save memory; tap Search all 4 lists to add the rest. To screen your own customers and keep a record of your decisions:

4. Open Customers. The first time, it asks for your name, which is stamped on your decisions.
5. Add a customer with reference CUST-001 and name Vladimir Putin, then click Onboard. It is screened right away. You can also import customers from a CSV or Excel file.
6. Open Review. Each possible match shows its list, score and strength (strong, possible or weak). Mark each one as a true match or a false positive and add a note.

How it works

Every day a small program downloads the official lists from the US Treasury, the EU, the UN and the UK government, puts them in one format, and signs them. When you open the app, your browser downloads the signed lists from aml-filter.com and checks the signature before using them. If the check fails, the app shows an error instead of quietly showing no matches. Names are then matched inside the browser tab by meaning (using a small language model that runs on your machine), by spelling, and by sound. Your customers and decisions are saved in your browser's own storage, not on a server.

While the app is open, it checks for a newer list every 30 minutes. When a list changes, it screens your saved customers again and only flags a match you already reviewed if something about it changed.

Which lists, and how fresh

List	Published by	Entries (25 Sep 2026)
OFAC SDN list	US Treasury	19,391
Consolidated list	European Union	6,241
Consolidated list	UN Security Council	1,011
UK Sanctions List (asset freezes only)	UK government (FCDO)	5,682

The lists are rebuilt daily. The app shows when each list was last updated. If a source could not be downloaded, the app says "Not updated for" and how long, rather than pretending the list is current. You can switch lists on or off, and set how strict the matching is, in Settings.

What it does not do

• It is not legal advice or a certified compliance product. It helps you find possible matches. A qualified person must confirm each one against the official source and decide what to do. See NOTICE.
• Sanctions lists only. It does not cover politically exposed persons (PEPs), news searches, or company ownership.
• One person, one browser. Customers and decisions live in the browser you used. There is no shared team view. Clearing your browser's site data deletes them. The spreadsheet export covers the customer table only, not your decisions.
• Needs the internet to open. There is no offline mode.
• Close misspellings, not wild ones. Vladimir Poutine still finds Putin, but as the second result. Read the whole list, not just the top line.
• Browsers: current desktop Chrome, Edge, Firefox and Safari 17 or later. Phones work in a lower-memory mode.

When to use something else

If you need	Use
Certified coverage, PEP and news data, and someone accountable	A commercial screening vendor
A few one-off checks	The official search pages, such as OFAC's Sanctions List Search
One shared review queue for a whole team	A screening server you host
Private checks across four lists with the reasons shown, and nothing to run	AML-Filter

Run it yourself

You need Node 22.13 and pnpm (which corepack provides).

git clone https://github.com/hseshadr/aml-filter
cd aml-filter/frontend
corepack enable
pnpm install
pnpm --filter aml-filter-app dev

Open http://localhost:5173/screen. Use localhost, not a network address, because the browser only allows the signature check there. The first run takes a few minutes while it downloads the name-matching model.

Your local copy searches a small made-up list of 8 people that ships with the code, so it works on a fresh clone. Type fakovic and it finds "Ivan Fakovich". For the real lists, use aml-filter.com. To host your own copy with real lists, see Deploy.

Develop

New to the code? Start with Getting started for developers. It takes you from a fresh clone to a passing local build and your first change.

cd frontend && pnpm gate

This runs the same checks as CI: lint, type checks, unit tests, builds, match-quality checks, and real-browser tests. It needs the exact Node version in frontend/.nvmrc.

More detail

• Getting started for developers: set up, run the checks, make a first change.
• Quickstart: a longer walkthrough of every page, for users.
• Architecture: the packages, how matching and scoring work, the security summary, and what the tests prove.
• Explore the interactive architecture map.
• Watchlist format: how the lists are signed and packaged.
• Match quality: how often the right name is found, measured on the real US list.
• Memory and browser support: how the app stays inside a phone's memory.
• Deploy: build settings and hosting your own copy.
• Operations: publishing lists and handling incidents.
• Security: how to report a vulnerability.
• Contributing: how we work and what a PR needs.
• Changelog: what changed in each release.

License

MIT. See LICENSE. Data attribution for the sanctions lists is in NOTICE.