Mask before you ask.

Everything you type into ChatGPT or Claude goes to someone else's computer, including the names, numbers, and addresses that slip in without you noticing. Veyl catches them before you hit send, masks them in your browser, and shows the real values back only to you.

Add to Chrome (free) Read a document safely →

Free · Works inside ChatGPT and Claude · Nothing leaves your browser unmasked

What slips out

You'd be surprised what rides along in a quick question.

A pasted email, a form, a paragraph from a document. AI chat is useful precisely because you give it real context, and real context is full of real information.

@

Names and contacts

Your client, your patient, your coworker, yourself. Names, emails, and phone numbers are the easiest things to paste and the hardest to take back.

#

Numbers that matter

SSNs, account and card numbers, medical record numbers, policy IDs, birthdays, home addresses. Veyl recognizes the shapes, even partly masked ones.

Whole documents

Contracts, statements, forms. One paste can carry a hundred identifiers. Veyl reads documents locally and masks them before any AI sees a word.

How it works

Four steps, and only one of them is yours.

Install the extension and keep using ChatGPT or Claude exactly as you do now. Veyl works quietly until it's needed.

1

You type, Veyl watches

As you write, Veyl detects sensitive information locally, using patterns plus an AI model that runs entirely on your device.

2

It asks before anything leaves

Hit send and Veyl steps in: "Sensitive information detected. Mask before you ask?" Review each item, keep or mask with one click.

3

Only the masked version goes

The AI receives [PERSON_1] and [SSN_1], never the real values. It answers just fine without them.

4

You see the real answer

Veyl restores the actual values in the reply, on your screen only, and keeps a tamper-evident log of what was protected.

Documents

Have a PDF or Word file? Let AI read it, veiled.

The Veyl workspace reads your document in your browser (even scanned ones, via local OCR), masks everything sensitive, sends only the masked text to the AI with your question, and restores the real values in the answer. Your file never gets uploaded anywhere. Bring your own AI key (OpenAI or Anthropic): your usage, your account, your control.

Open the workspace →
Questions

Fair questions, straight answers.

Where does my data actually go?
Detection, masking, and restoration happen inside your browser. Veyl has no server that sees your text. We could not read it if we wanted to. The only thing that leaves is the masked version, sent to the AI site you were already using. The one-time download of the local AI model is a download, not an upload.
What does it catch?
Names, organizations, and places (via a local AI model), plus pattern-based identifiers: SSNs and ITINs (even partial ones like XXX-XX-1234), payment cards, account numbers, medical record numbers, dates of birth, phone numbers, emails, street addresses, policy and reference numbers, and tax preparer IDs. Plus anything you add to your personal always-mask list.
What if it misses something?
It will, sometimes. No detector is perfect, and we say so plainly. That's why you review before sending: every detected item shows as a pill you can toggle, and you can select any text it missed and mask it in one click. You always have the last word, in both directions.
Is it really free?
The extension is free. The document workspace is free too: you bring your own AI key (OpenAI or Anthropic), so your AI usage is billed to your own account at their rates, not through us. We don't run ads, and we don't sell data. There is no data of yours here to sell.
Does it work with Gemini? On mobile?
Today: ChatGPT and Claude, in desktop Chrome and Edge. Gemini support is next on the list. Mobile browsers don't support extensions, so phones are a longer-term story.