Introducing Safe Zone: how safe is your AI chat, really?

Most AI apps don't really tell you where your data goes. You sign in, you type, you get an answer. Did your chat stay in Australia? Did anyone outside the company keep a copy? Has it been used to train someone else's model? Mostly, you don't know.

Safe Zone is our answer to that.

It's a little badge at the top of Opti Assist. Look up. There it is. It tells you, right now, how safe what you're doing is on a scale of 1 to 5. Hover it and you get the full story: what level you're at, what's protecting you, and the one thing that would make it stronger.

The whole thing is built around one simple question: how safe is what I'm doing right now?

The five levels

The badge goes from Level 1 (Secured) to Level 5 (Fortified). It always picks the highest level your live chat actually qualifies for.

Level 1, Secured. You're signed in with your work or school Microsoft account, and every action is recorded. Identity is sorted, actions are logged. If you're chatting with an outside AI (like ChatGPT or Claude on their own apps), that bit might be kept by them for a while. The badge tells you that.

Level 2, Sovereign. You're using an ORCA model. Aussie-built, Aussie-hosted, not kept by anyone outside. Your chat stays in Australia and inside ORCA. Nothing leaves. This is the everyday resting place for most people using ORCA.

Level 3, Guarded. Sovereign chat with ORCA Opti AI Guardian on at its Essential level. On top of sovereignty, Opti AI Guardian is reading along. It hides personal info, blocks dodgy questions and stops the common AI tricks. This is the first level where the chat itself is being watched over, not just held in a safe spot.

Level 4, Hardened. Sovereign chat with AI Guardian set to Enhanced. The trickier AI attacks that the Essential level might miss? Caught here.

Level 5, Fortified. Sovereign chat with AI Guardian set to Maximum. Everything we know how to block is being blocked. The strongest setting on the standard scale.

If you're counting, three things change the level: who you're signed in as, which AI is actually running the chat, and whether AI Guardian is on (and how strongly).

We show you the worst, not the best

The most important word here is honest.

The badge always shows the weakest part of what's happening right now, not the best feature you happen to own. If your team has AI Guardian set to Maximum but you're mid-chat with ChatGPT directly, the badge says Level 1. Because that's where you actually are.

A reassuring badge while your data was walking out the door would be the opposite of helpful. And trust is the whole point.

It's the same reason we don't paint the lower levels red. Being on a lower level isn't a fail. Level 1 is still safe on identity and audit. You're signed in. You're audited. It just means the AI chat itself isn't sovereign yet. We never go red. We tell you, clearly, what's going on. We don't try to scare you.

Where your data lives matters most

The five levels aren't five independent switches. There's an order to them.

Where your chat goes, and whether anyone outside ORCA keeps a copy of it, is the biggest thing. You can't reach the top three levels if your chat is leaving for somewhere else. No amount of extra protection makes up for that.

That's why AI Guardian only kicks the level up when you're on a sovereign chat. AI Guardian looks after what happens inside the chat. Sovereignty looks after where the chat lives. Both matter. But they have to stack in the right order.

What sits above and beside

Two bigger versions of ORCA sit above the everyday Lite / Standard / Pro models, for organisations that need more.

• ORCA-Private. Your own dedicated kit in an Australian data centre. Not just your own corner of ours, your own room. Talk to us.
• ORCA-Classified. Defence-grade infrastructure, accredited for classified material from PROTECTED through TOP SECRET. Aligned to the ISM and the DISP. Air-gapped, secure-room, and cleared-people-only options if you need them.

Both are talk-to-us, not sign-up-now.

There's also a spot on the scale we've kept aside, called Gated (or Enclave), for a future ORCA model that lives outside Australia. No model fills it today, so the level isn't live yet. And there's a premium setting above Fortified, called Defence-Grade, paired with ORCA-Classified. It's set up and certified per organisation rather than calculated automatically, because a lot of what makes it work happens in the contract and the infrastructure, not just in the session.

Where to look

Once you're signed into Opti Assist, the badge is at the top of the app, right next to your team name. Hover it for the checklist and the suggestion. Try switching between an ORCA model and an outside one, and watch what happens.

The full version, with all the levels, the principles behind them and a longer FAQ, lives on the Safe Zone page. If you'd like to see your own Safe Zone level, sign in with your work or school Microsoft account at portal.orcaopti.ai and the badge will tell you, in plain English, exactly where you are.