Veritus

Monitoring boundaries and data handling

What Veritus collects, and what it doesn't

Veritus records which AI tools are used and what your policies decided. It is not a keylogger, a screen recorder or a general employee-monitoring tool. This page lists exactly what each part collects, where it is stored, who can see it and how long it is kept, so you can explain it to your clients before you deploy.

Browser extension for Chrome and Edge

Collects

  • Only on six AI sites: ChatGPT, Claude, Gemini, Copilot on the web, Perplexity and DeepSeek
  • The prompt text and readable text extracted from supported file uploads, sent to Veritus to check against your policies
  • The decision, what was detected, and what the person did next (cancelled, sent with a reason, or blocked)
  • Which enrolled person it was, the AI site address, browser and operating-system version, and the extension's health
  • When an AI site your organization has not approved is blocked: the site name, at most once a minute, and any access request with the reason the person gives

Never collects

  • Anything on other websites, and browsing history
  • AI responses
  • Screenshots, keystrokes outside the prompt box, or files that are not uploaded to an AI site

Stored prompts are redacted by default: identifiers such as email addresses are replaced with labels and only the first 220 characters are kept. When sensitive data is detected, the stored text is replaced entirely. An organization can choose to store full prompts instead. This protects the stored record. Automatic removal of sensitive text from a message before it reaches an AI tool is not enabled in this release. Use a blocking policy to stop sensitive content from being sent.

File checks depend on the installed extension, file type, size and whether readable text can be extracted. Unsupported formats, oversized files and unreadable content are not fully inspected. Browser policies require an enrolled, connected extension; a computer appearing in the inventory does not mean its browser prompts are being checked.

Windows agent (observe only)

Collects

  • Names, publishers and versions of installed applications
  • Names of programs that start or run, the program that started them, and their file fingerprints (SHA-256), including system tools such as PowerShell
  • Network destinations those programs connect to, and DNS lookups
  • Names of MCP servers set up in AI apps (Claude Desktop, Cursor, VS Code, Windsurf, Claude Code), with the program or package they run, and AI-related browser extensions by name and version
  • Computer name, Windows version and the agent's own health
  • When agent controls close or flag an app: which app and when; when someone asks for access from the notice: their Windows account name and the reason they give

Never collects

  • Prompts, AI responses or anything typed or shown inside an app
  • Documents, file contents, screenshots or keystrokes
  • Command lines, window titles, or file paths (paths are off by default)
  • Settings, keys, arguments or addresses inside AI app configuration files
  • Who is signed in to the computer

The agent records every program, not only AI tools, because it has to see a program to tell whether it uses AI. That is why process names such as PowerShell or Windows sign-in can appear. Veritus records that they ran, not what they did. The console shows AI-related activity first; the full record is one click away for troubleshooting. The agent closes desktop apps marked Not approved only where the organization turns on agent controls, and tells the person why.

How much is kept

Each organization's MSP chooses how much detail Veritus keeps. Minimal keeps installed applications and AI activity. Standard, the default, also keeps the first record of each program and destination per computer each day and counts the repeats. Detailed keeps every record for up to 7 days during an investigation, then switches itself off. Whatever the level, each computer keeps at most 20,000 records a day, with room kept for AI activity.

What Veritus can see about AI use

Where AI is usedWhat Veritus seesCoverage
AI websites in Chrome or Edge, with the extensionPrompt and upload content is checked against your policies, and each decision is recorded.Content checked
Desktop AI apps such as the Copilot, ChatGPT or Claude appsThat the app ran, when, on which computer, and which AI services it contacted. Not what was typed or shared.Use only
AI features inside other apps, such as Microsoft 365 Copilot in WordNot covered today.Not covered
Firefox, Safari, macOS and mobile devicesNot covered today.Not covered

Policies act on content, so they apply where content is checked: AI websites with the extension. For a desktop AI app, Veritus can show that it is in use so you can review it, steer people to the governed web version, or block it with your endpoint management tool.

Where data is stored

  • In the Veritus service database, on servers operated by Veritus. Data is not sold or shared with third parties.
  • Ask your Veritus contact for the current hosting provider and region.
  • Everything travels over HTTPS. The Windows agent refuses to send data over an unencrypted connection.
  • Enrollment codes are encrypted, sign-in passwords are hashed, and every organization's data is kept separate by the application.

Who can see it

MSP administratorsEvery organization their MSP manages.
Organization administratorsTheir own organization only.
EmployeesNo console access. They see only the extension's warnings and blocks.
Veritus staffOnly through an audited support mode, recorded in the audit log.

Every administrator action, including revealing an enrollment code, is recorded in the organization's audit log.

How long it is kept

Windows agent recordsKept for 30 days, then deleted within the hour.
Daily counts per program and destinationKept for 90 days. Repeats of the same program or destination are counted here rather than stored one by one.
Agent records waiting to uploadHeld on the computer for at most 48 hours if it is offline.
Browser policy decisions and prompt excerptsKept until the person or the organization is deleted. Activity shows the last 90 days; reports can cover older periods.
Administrator audit logKept until the organization is deleted.
After deletionData is removed from the live database. It can remain in Veritus server backups until those backups are replaced.

Deleting a person deletes their recorded activity. Deleting an organization deletes all of its data. Uninstalling the Windows agent can also remove its local data.