GovernedGoverned AI at the deployment layer.
Frontier labs govern what a model can do. The Cloud governs what an agent does with your data, on your behalf, right now. That second layer — the deployment layer — is where an AI actually touches your email, your files, your calendar, the people you'd share with. We govern it.
Two layers of AI safety
Capability safety is the lab's job — teaching a model what it will and won't generate. Deployment safety is ours — governing what the model is allowed to do once it's inside your workspace, acting on your real data. A frontier lab makes the model safer. A governed workspace makes its use safer. Both matter, and they are different layers.
How the deployment layer is governed
- AI that asks before it acts.
Everything you make starts private. Sharing something, changing who can see it, or any action it can't undo needs your consent for that specific action — every time. Ora asks before it opens a single thing to anyone, and never infers a share from a broad instruction. This is the consent floor: not a setting you find, but how The Cloud is built.
- You work with Ora, one governed identity.
On The Cloud the AI is Ora, and it always presents as Ora — one accountable collaborator you delegate to, not whichever model is underneath. It stays transparent about the engine it runs on when you ask; the identity is governed, the model is disclosed, never hidden.
- Your workspace is yours alone.
Every read is bounded by your tenant key, so one person's memory, files, and work never cross into another's. Cross-tenant access fails closed — a hard rule enforced at the data layer, not an option you switch on.
- What the AI touched stays on the record.
When Ora helps write something, that's kept with the work: which model did it, and what kind of change it made. The trail of what the AI did travels with the result, so its hand in your work is legible after the fact.
- You only see what's finished.
While Ora writes to a page, a half-formed link never renders as something you can click — nothing mid-stream pretends to be done. The safe default holds even as the answer is still arriving.
What a governed workspace is
A governed workspace is a place where AI can act on your real data, and every action runs inside rules you set: consent before it shares, one accountable identity, your data isolated to you, a record of what the AI did. The model is governed at the lab; the deployment is governed here. Nobody else owns the governed workspace — so we're building it.
An open OS that is also governed — the safe place to let agents act. You arrive already inside a working WorkSpace; the consent floor, the isolation, and Ora's governed identity are there from the first keystroke.
The Cloud is a lab and a campus, working. We won't sell you and we won't beg you. What we've built is here — the AI serves the person, the person governs the AI — and the door is open. Come in and put an agent to work.