Introduction

Vloex is an AI usage monitoring and governance platform built by Gofylo Technologies. This privacy policy explains what data the Vloex browser extension and platform collect, how that data is used, and the controls available to organization administrators.

Vloex is designed for enterprise deployment. Data captured by the extension is sent to your organization's Vloex backend — not to a shared multi-tenant service. Your organization's administrator controls what is captured, and can delete captured data on request. There is no administrator-configurable retention period today — see Data retention below.

What the extension collects

When installed and configured, the Vloex browser extension captures data about AI tool interactions on supported provider websites. The scope of data collected depends on the coverage level configured by your organization's administrator:

Full trace — provider name, model, timestamps, estimated cost, the prompt text, the response text, the prompt token count and the response token count are all captured. Nothing is stripped.
Chat only — the prompt text and the prompt token count are captured, along with provider name, model, timestamps and estimated cost; the response text and the response token count are not retained.
Metadata only — provider name, model, timestamps and estimated cost are captured; the prompt text, the response text, the prompt token count and the response token count are all dropped.

Text a level excludes is removed on your device before the event is written to the extension's local queue, and removed again by the backend when the event arrives — the server applies your organization's level itself rather than trusting what the device sent. Removal is one-way: text already dropped cannot be recovered if an administrator later loosens the level.

A coverage level governs what is retained, not what is transmitted for the pre-send check. When enforcement is enabled, your prompt text is sent to the policy-evaluation endpoint before the request leaves the browser at every level, including Metadata only. That call returns a decision and stores nothing.

For all coverage levels, the extension also collects:

Chrome profile email address (for identity resolution)
A device identifier (for enrollment tracking)
Sensitive data detection results — the pattern types found, never the matched text. These are derived by the backend from the text a coverage level allowed through, so Metadata only records none at all, and Chat only records them for prompts and not for responses.
The name, size and type of any file you attach to an AI prompt (not the file itself — see Sensitive data detection)

What we do NOT collect

Your browsing history — the extension requests no tabs or history permission. It sees only the pages it is injected into, listed below.
Saved passwords or login credentials — no credential store, no autofill, no login form is read. A credential typed or pasted into an AI prompt is captured with the rest of that prompt, which is precisely what sensitive-data detection exists to flag.
Files on your device — the only files read are the ones you attach to an AI prompt. Their name, size and type are recorded, and text-extractable files up to 1.5 MB are read so their contents can be scanned before the prompt is sent. Nothing else on disk is opened, and downloads are never read.
Keystrokes — there is no keylogging, on AI sites or anywhere else. What is captured is the request an AI site sends, not the typing.
Screen captures or screenshots — never, on any site.
Page content from sites that are not AI providers — with one exception, named here rather than buried: a read-only script runs on Gmail and Outlook and reads the signed-in email address and nothing else. See Browser extension permissions below.

Sensitive data detection

The extension scans prompts for sensitive data patterns (such as Social Security numbers, API keys, credentials, and personally identifiable information) before they are sent to AI providers. The pre-send check sends the prompt text to the Vloex backend to be scanned; that request is not stored, and the response reports only the types of patterns found and their counts — never the matched values. If the backend cannot be reached in time, the extension falls back to scanning locally in the browser.

Files you attach to a prompt are scanned the same way. A text-extractable file up to 1.5 MB is read in the browser and its contents sent to be scanned; the contents are not stored, and the event records only the file's name, size, type and what was detected. Larger or binary files (PDF, Office, images) are not read at all, and the event records them as unscanned rather than letting them pass as clean.

When policy enforcement is enabled, the extension can block, redact, or warn before sensitive data is sent to an AI provider.

Data storage and encryption

In transit — All data is transmitted over HTTPS (TLS 1.2+).
At rest — Prompt and response text is encrypted using AES-256-GCM before storage.
Offline queue — Events captured while the backend is unreachable are stored temporarily in the browser's local storage and transmitted when connectivity is restored.

Data retention

Captured data is retained until it is deleted. There is currently no automated retention or purge schedule, and no administrator-configurable retention period — a scheduled purge is not implemented. Deletion is performed on request and runs immediately: it removes the captured interactions, prompt and response text included. It does not remove everything. Enforcement records, the audit log and the audit-chain anchors are kept — the anchors for seven years by default, because an evidence chain with holes in it cannot be verified — and so is the inventory of AI apps discovered in your workspace.

Third-party data sharing

Vloex does not sell, rent, or share captured data with any third parties. Data is sent only to your organization's Vloex backend instance. There is no advertising, no analytics tracking of end users, and no data broker relationships.

Administrator controls

Organization administrators have full control over:

Coverage level (full trace, chat only, or metadata only)
Which AI providers are monitored
Policy enforcement rules (block, warn, redact, allow)
Which users and departments are enrolled
Export of captured data

Retention period is not among these controls. There is no setting that expires data after a chosen number of days; deletion is an action an administrator takes, not a schedule they configure. Deletion is also not a control in the dashboard today — it is an API action, and Data retention above says what it removes and what it keeps.

Browser extension permissions

The extension requests only the permissions necessary for its function:

storage — Stores preferences and an offline event queue.
alarms — Retries failed event submissions on a schedule.
identity / identity.email — Retrieves the Chrome profile email for automatic enrollment.
Host permissions — The shipped extension declares 171 host permissions and injects its capture script into 166 AI assistant sites, generated from our provider registry rather than a hand-kept shortlist of four. The remainder are your Vloex backend and the Google and Microsoft APIs used for workspace discovery. The exact list ships inside the package and is readable at any time at chrome://extensions → Vloex → Details → Site access.
Gmail and Outlook — A separate read-only script runs on mail.google.com, outlook.office.com and outlook.office365.com. It reads exactly one thing from the page, the signed-in email address, so that a device can be attributed to a person without asking them to type it. It reads no message content, and sends nothing else.

Changes to this policy

We may update this policy as the product evolves. Material changes will be communicated to organization administrators. The effective date at the top of this page reflects the most recent revision.

Contact

For privacy questions or data requests, contact us at sats@vloex.com.

Gofylo Technologies

Privacy Policy