Using Zeuslock on ChatGPT, Claude, Gemini and Copilot

A quick tour of how Zeuslock sits beside you in your favourite AI tools — toolbar icon, inline indicator, pre-send preview and what to do if you're blocked.

Welcome — Zeuslock runs alongside you

Zeuslock is the small shield that quietly watches the prompts you send to your AI tools. It works out of the box on ChatGPT, Claude, Gemini, Copilot, Perplexity, Mistral, Poe and You.com, and there's a Universal mode that covers any other AI site your team uses. You don't need to learn a new interface — just keep working the way you already do. Zeuslock only steps in when something sensitive is about to leave your machine.

This page walks you through the four things you'll actually see during the day: the toolbar icon, the inline indicator next to the prompt box, the pre-send preview, and what happens if a prompt is blocked.

The toolbar icon

The small Zeuslock shield in your browser toolbar tells you, at a glance, whether you're protected and whether anything needs your attention on the page you're on.

  • Green — connected to your organisation and actively monitoring this page. Nothing to do.
  • Amber — a warning is visible on the current page. Usually means a finding was detected in what you've typed; open the indicator next to your prompt to see what.
  • Red — a block is in effect. Your last send was stopped, or your security team has set this site to block-only mode. The page will tell you what happened.
  • Grey — not connected to your organisation. Sign in again from the toolbar, or contact IT if it stays grey.

Click the icon any time to see the current site, the active policy mode, and a shortcut to your incident history.

The inline indicator next to the prompt input

On every supported AI tool, Zeuslock adds a tiny dot at the edge of the prompt box. It changes colour in real time as you type or paste:

  • Grey dot — nothing sensitive detected yet.
  • Amber dot with a counter (for example, 3 findings) — Zeuslock has spotted things like an API key, an email address or a credit-card number. You can still send, but the items will be handled according to your organisation's policy.
  • Red dot — at least one item will block the send. You'll need to remove or rewrite that part.

Hover the dot for a quick list — the types found, where in your prompt they appear, and what the policy will do (Monitor, Anonymize or Block). Click it to open the full preview.

The pre-send preview (Monitor and Anonymize modes)

When your organisation runs in Monitor or Anonymize mode, Zeuslock shows you a quick preview of what your prompt will actually look like once it leaves your machine.

  • In Monitor mode, your prompt is sent unchanged but the finding is logged for your security team. The preview is mostly there so you know what was detected.
  • In Anonymize mode, sensitive items are replaced with format-preserving placeholders — a real credit-card number becomes a structurally valid fake, so the AI still understands the shape of your data.

Press Confirm to send the redacted version, or Cancel to go back and edit the original. You can also turn off the preview for the current session if you trust what you're sending.

What happens when you're blocked

If a prompt contains something your security team has set to Block — typically credentials, private keys, customer records or proprietary source code — Zeuslock stops the send before it leaves your browser. A friendly modal explains:

  • What was found — for example, "AWS access key (AKIA…)" or "customer email address".
  • Why it's blocked — a one-line plain-language reason tied to the policy.
  • How to rewrite safely — suggestions like removing the value, replacing it with a placeholder, or describing the problem without the secret.
  • Request an exception — if you genuinely need to send it, a button lets you ping your security team with the context.

You won't lose your draft. The modal closes back to your prompt box with everything still there, so you can edit and try again.

Your privacy

Zeuslock detects sensitive content locally, in your browser. Only redacted metadata — the type of finding, a length, a hash — ever leaves your machine. The original sensitive content stays with you. The Zeuslock backend is hosted in the EU and processes nothing more than what's needed to enforce your organisation's policy.

Performance — you shouldn't feel it

Zeuslock adds roughly 50 to 100 milliseconds between you hitting Send and the prompt reaching the AI tool. That's faster than the network round-trip itself. In day-to-day use you won't notice any lag, even on long prompts or pasted documents.

Frequently asked questions

Will Zeuslock read my chats?

No. Zeuslock only sees what you actively type or paste into the prompt box. Past conversations, the AI's responses and anything else on the page are never inspected or stored.

Does it work on the mobile site?

Yes, on any mobile browser where the extension is installed (Firefox on Android, for example). For native mobile apps, where browser extensions can't run, your organisation uses the desktop agent on tethered laptops instead.

What if I disagree with a block?

Use the Request exception button in the block modal. It opens a short form that goes straight to your security team with the page, the policy and your reason. They'll usually reply within a few hours.

That's everything you need for day-to-day use. If anything looks wrong — the icon stays grey, a site isn't being monitored, or a block doesn't make sense — open the toolbar and use the Report a problem link to send the details to your IT team.