Documentation

Documentation

YuviRed on your work computer: a guide for the people who use it

Your organization put YuviRed software on this computer to see how AI tools are used at work and to keep sensitive information from being sent to AI services by mistake. This page explains what is on your computer and what it does. It does not replace the notice your organization gave you.

What runs on your computer

  1. The YuviRed Endpoint Agent. A background Windows service named "YuviRed Endpoint Agent". It starts with Windows, runs as the system, and reports about once a minute to your organization's own YuviRed server. It does not report to YuviRed LLC, the company that makes the software.
  2. The YuviRed AI Prompt Monitor (only if your organization loaded it into your browser; in this version it is added by hand, browser profile by browser profile, not installed automatically). A Chrome or Edge extension that works on AI chat sites such as ChatGPT, Claude, Gemini and Copilot.

What the agent reports

  • Your computer name, your Windows user name, one IP address and the Windows version.
  • Which AI programs are running (name, location, memory and processor use).
  • Web connections of programs to web ports, with the address, its name and the program. This covers every program, not only AI. The amount of data is not recorded.
  • Visits to AI websites in Chrome, Edge, Brave and Firefox: a count and the last time, for the last 24 hours. The agent reads a temporary copy of your browser history on the computer to find these. Other sites, page addresses and page titles are not sent.
  • AI-related browser extensions, installed AI software, AI services running on the computer, and AI entries in the Windows name cache.
  • Settings and names of AI tools such as coding assistants (secrets are removed; environment variable names, never values).

What is not collected

Keystrokes, screenshots, the clipboard, passwords, the contents of your documents, the contents of web pages, and chat transcripts. The agent does not read the text of your prompts or the AI's replies.

What the browser extension does

On 15 AI sites (ChatGPT, Claude, Gemini, Microsoft Copilot, Perplexity, Microsoft 365 Copilot, DeepSeek, Grok, Le Chat, Meta AI, Poe, Google AI Studio and HuggingChat) it checks what you type, paste or attach, and what the AI replies, for sensitive information (passwords, keys, personal data, card numbers). The check of what you send runs before it is sent, and it also looks at a draft while you type. The check runs in your browser. Your organization sees the kind of finding, a masked fragment (at most the first 4 and last 4 characters; for a password only the beginning), what happened, the site, and your browser. It does not send the text. Only if your organization has switched on full capture, which the notice must tell you, is the first part of a flagged prompt or reply stored on its server, encrypted, for 24 hours.

In "monitor only" mode you see nothing on the page, but findings are still recorded. Nothing is recorded or reported until consent is given. In this version that means you tick the box **I consent to local inspection of prompts and attachments on the monitored sites above** on the extension's options page and choose Save; the page opens by itself when the extension is loaded and consent has not been given. If your organization ever sets consent through browser policy instead, the same page shows "Policy consent value: true (consent given by policy)" and you are not asked.

If a warning appears

A box titled Potentially sensitive information detected says what kind of information was found and asks "Are you sure you want to send this information to <the AI service>?".

  • Choose Go back and edit (the default) to take the information out. Nothing is sent.
  • Choose Send anyway only if you are allowed to share it. Your choice is recorded. Closing the box or pressing Escape does not send.

If a block appears

A box titled Prompt blocked: sensitive information detected lists what to remove. Remove it and send again. You cannot override a block. A small "ref" code at the bottom of a warning or block lets your administrator find the event if you ask for help.

If a draft notice appears

While you type, a small notice may say "Sensitive information detected in your draft" and that your draft has not been sent. It has a close (x) button. Nothing is sent by the notice itself; it tells you what will happen when you press Send.

The options page

Open it from the browser's extensions page, Details, Extension options. It shows what is inspected, where reports go, and a Connection status ("Connected to YuviRed Lite" when it works). A Copy status for support button gives your administrator version and status numbers, with no prompt text.

Where to read the notice

  • On this computer: C:\Program Files\YuviRed Agent\NOTICE.txt.
  • On your organization's server: https://<your server>:8443/static/legal/notice.html. Ask your administrator for the address. The data notice (data-notice.html) gives more detail.

What you cannot change

You cannot stop, pause or remove the agent without administrator rights, and you cannot change your organization's policy or the rules. A computer whose agent stops reporting shows as Stale and then Offline on your organization's console, and a browser whose extension stops reporting shows as Stale there. The extension's options page lets you narrow the list of monitored sites or lower its capture level; your organization's console is not told when you do. If the browser extension is ever installed by browser policy (not in this version) it has no Remove button and cannot be turned off.

The agent's Windows service is named SyntinelRedAgent and its files are under C:\ProgramData\SyntinelRed (the product's earlier name; it will change in a later release).

Questions and requests

Contact your organization's privacy contact (shown on its notice page, or ask your administrator). You can ask what is held about you, to correct or delete it, or to object, as the law where you work allows. YuviRed LLC does not hold your monitoring data; its address for software questions is support@yuvired.com.