FAQ about the extension

From the Chrome Web Store, with no ZIP file and no developer mode: click the install button on the home page, then "Add to Chrome". In Edge, first allow extensions from other stores once. This takes about a minute; the steps with screenshots are under Install in Chrome and Install in Edge.

Automatically. The browser fetches new versions from the Chrome Web Store by itself; you don't need to do anything. What has changed is listed under Changes.

Nothing, for individuals and for companies. The extension needs no account and no subscription, and it is open source: the code is on GitHub (GNU AGPL v3).

Not through AImori. The check runs locally in your browser, the extension has no network access and stores no input. Only counters without content (daily counts for 90 days and, if enabled, monthly counts for the AI literacy record for 13 months) and your settings are stored.

AImori never sends anything itself. The bar while typing is just a hint on your device. Whether an AI site already reads text before it's sent, for example for drafts, is up to the site itself – so it's best not to type sensitive data in the first place.

Yes. The detection rules, the local AI model and the text recognition for images are part of the extension package. The check needs no connection, and the extension is not allowed to open one.

Five fixed permissions without access to other websites: "storage" (settings and counters), "scripting", "offscreen", "activeTab" and "declarativeContent" – what each is for is explained under Privacy. It checks input only on the supported AI sites, with no access to history or downloads. Optional and only after your click: further AI sites from the list (for a notice or block on AI sites that are not approved), confidential sources and form sites set by your organisation, and all https sites for the copy notice; the browser asks clearly for this, and you can revoke it at any time.

242 data types in 60 packs: IBANs and credit cards, national ID, social security, tax and identity numbers of many countries, passport and ID card lines, dates of birth, contact data, API keys, passwords and tokens – also inside pasted code – as well as health data and customer lists in context. Where a check digit exists, it is validated. Examples with test values: What is detected.

Yes, with the local AI model: names of people, companies and organisations, postal addresses and health data. The model runs on your device, without network. You can turn it off in the settings; then only the fixed rules apply.

Images, yes: pasted or uploaded screenshots (PNG, JPEG, WebP) are read locally with text recognition, and if something is found, AImori offers a redacted copy. Tables (CSV, XLSX) and text files are checked for customer lists, hidden instructions and – if your organisation has set this up – parts of company documents. PDF or Word files are not read; a notice appears instead. Organisations can block file uploads on AI sites via policy.

Choose "Not sensitive" in the dialog and confirm with the tick box – AImori then no longer reports that value for 30 days. This only works for data types set to "Warn", only if your organisation allows it, and not for phone numbers and IP addresses. To change a whole data type, set it to "Warn" or "Off" in the settings. Because numbers only count with a valid check digit, false alarms are rare. Feel free to report them with a made-up example on GitHub.

"Redact" replaces the value with its kind, for example [Financial data]. "Placeholders" replaces it with a token like [IBAN_1]; the same value always gets the same token. You can then read the AI's answer with the real values inside the extension – the AI itself never sees them (Warning bar and actions).

Yes, if your organisation hasn't turned it off: click Send anyway in the bar, then click Really send? a second time within 4 seconds. AImori then sends the unchanged text.

Microsoft does not let extensions run on copilot.microsoft.com in Edge, including force-installed ones. In Google Chrome, Copilot is protected.

Yes, via Group Policy, Intune or a registry script with the extension ID lefajbhockjgdhaohnacgaggfepianjh, with fixed rules per data type, allowed and blocked AI sites and, if wanted, under your own name and logo. The organisation sees no content in the process (For companies).