Veil Buy Veil

Don't take our word for it. Check.

No telemetry. No analytics. No accounts. No licence checks. The Veil app sends nothing anywhere except your requests to the AI provider you chose, and only after sensitive values have been replaced.

Verify it yourself

The engine that redacts your data is open source. The rest of Veil we made easy to observe.

01

Watch the network

Veil listens only on your own computer (127.0.0.1). Open your system network monitor and see for yourself which connections it makes.

02

Use the test playground

Paste text and see exactly what the AI provider would receive. Nothing in the playground is sent anywhere.

03

Read the activity log

Every request is listed with what was replaced, by type and count. Your content is not stored in the log.

04

Read the engine's code

The engine that redacts your data and the browser extension are public on GitHub, and Veil shows which versions you run.

How to check it
Every network call Veil makes
AI provider
Your requests, after replacement, go to api.anthropic.com or api.openai.com, the provider you chose.
Each request
Name detection
Only if you turn it on: downloads the model from huggingface.co and its runtime from pypi.org. Sends nothing else.
Once
Update check
The app doesn't check. The Chrome extension's popup reads getveil.tech/download/latest.json, at most every 6 hours, to say when a newer version is out. The request carries nothing about you.
When you open the popup
Licence
None. Your licence key is verified on your computer.
Never
Chrome extension
Passes your messages and attached files to the Veil app on your computer, through Chrome's native messaging. Its code is public.
Each message

What Veil can and can't protect

Privacy software should be clear about its limits.

It protects

  • Emails, phone numbers, card numbers, bank accounts, keys and secrets, IP addresses, web addresses and national IDs, in the text you send
  • PDFs you attach, in the desktop apps and on claude.ai and chatgpt.com: Veil reads their text and replaces the same values in it
  • Text files you attach on claude.ai and chatgpt.com, such as notes, Markdown, CSV, JSON, code and .env files, in any common encoding
  • Names of people, companies and places, found by optional name detection that runs on your computer, and the names, client names and codenames you add to your own list
  • Streaming replies and tool calls, so tools still receive the real values they need
  • The same value always gets the same placeholder, so the AI can still follow who is who
  • If Veil hits an error, it blocks the request instead of sending it unprotected

It does not protect

  • The mobile apps, and chat websites other than claude.ai and chatgpt.com, such as Gemini or Copilot
  • Images, and files other than PDFs and text files, that you upload on claude.ai or chatgpt.com. Veil blocks them or lets them through, as set in Behavior
  • Text files a website makes by itself, rather than ones you attach
  • Images. Veil does not look inside them
  • Files other than PDFs and text files. Veil can read and redact those, but not zip files, Word documents, spreadsheets or presentations
  • Names that name detection misses and that are not on your list
  • Anything you send through a tool that is not pointed at Veil

Where your data lives

The vault remembers which placeholder stands for which value, so replies can be restored.

  • The vault is encrypted on your disk with AES-256-GCM.
  • The key is held in your system keychain, not in a file next to it.
  • Entries expire automatically after a period you choose, 7 days by default.
  • A memory-only mode keeps nothing on disk at all.