Browser extension privacy
This page covers the Itzal browser extension specifically, as required by the Chrome Web Store. It is separate from the website's privacy page, which describes something else entirely.
Draft, not yet in force. This describes what the extension actually does, written by the people who built it so that a review starts from facts rather than a template. It is not legal advice and has not been reviewed by counsel.
The extension reads the text you are about to send in a chat box on a small, named list of AI websites, and passes it to the Itzal agent running on the same computer. The agent decides whether it contains sensitive data and answers in a fraction of a second. The extension sends nothing over the internet. It has no servers to send anything to.
Itzal is deployed by an employer to its own staff. Your employer installs it, sets the policy, and receives the records it produces. Itzal the company does not receive your data. The agent uploads to a console your employer controls; where that console runs is their decision, not ours.
For data-protection purposes your employer is the controller and Itzal is a processor acting on their instructions. Questions about what is retained, who can see it, or why it was deployed are questions for them, and they are entitled to answer them.
It does not read any other page, does not watch your browsing, does not record what you type outside those message boxes, and has no access to other tabs.
The text is handed to the Itzal agent over Chrome's native messaging: a channel between two programs on your own computer. It does not cross a network. The agent runs the detection locally and replies with a verdict.
If the text matches your organisation's policy you see a warning on your own screen, and depending on that policy you may be able to continue anyway. The decision is shown to you before anything is sent.
The extension transmits nothing. The agent transmits a record of what happened to your employer's console, containing:
***-**-6789;The text itself never leaves the device. That is not a promise about our intentions: it is why the detection runs on your computer instead of ours. There is no copy of what you typed on any Itzal system, because it is never sent to one.
Two things, both on your own computer and neither containing any content:
These are numbers and yes/no flags. No text, no addresses, no identifiers.
chatgpt.com, chat.openai.com, claude.ai, gemini.google.com, chat.deepseek.com, perplexity.ai, copilot.microsoft.com, m365.cloud.microsoft, grok.com, chat.mistral.ai, meta.ai and poe.com.
That list is the coverage. The extension does nothing anywhere else, and an organisation deploying it should know which tools its people actually use.
nativeMessaging | Reaching the agent on your computer. Without it there is no detection at all. |
storage | The two content-free records above. |
scripting | Attaching to the message box on the sites listed. |
| Site access | Those sites and no others. There is no “all sites” permission. |
None. The extension contains no analytics, no advertising, no trackers and no third-party code. It makes no network requests of any kind. Nothing is sold, and nothing is shared with anyone other than your own employer's console.
The extension keeps its counters for a rolling window and overwrites them. Everything else (how long the records in the console are kept, and who may read them) is set by your employer, subject to a ceiling in their contract with us.
Itzal is workplace software, licensed to organisations and installed on equipment they manage. It is not directed at children and is not offered to consumers.
If what the extension does changes, this page changes with it before the version ships. Material changes are described in the release notes rather than appearing quietly here.
Questions about this page, or a request about your own data: privacy@itzal.ai. If you are an employee of a customer, your own employer can answer most questions faster, because the records are theirs.