Hybrid edge and sovereign AI without vendor lock in

The false choice between power and control

For years the market has offered organisations a single, uncomfortable bargain. If you wanted capable AI, you sent your data to someone else's cloud, accepted their terms, and trusted that the round trip would never come back to haunt you. If you wanted control, you settled for something smaller, slower, and easier to reason about. We built Mickai because that trade is not a law of nature. It is a design decision, and it can be reversed.

Mickai is a Sovereign Intelligence Operating System, a SIOS. It runs on the customer's own hardware. It can run entirely on premises, and it can run air gapped, with zero data egress and no public cloud round trip. The point of hybrid edge is not to be clever about topology. The point is to keep the sensitive work close, keep the control local, and still get the performance that used to be locked behind someone else's data centre.

What hybrid edge actually means here

Hybrid edge is often sold as a marketing phrase that means very little. We use it precisely. It means the heavy, sensitive inference happens on the metal you already own, in your building, under your physical control. Orchestration, governance, and the coordination of many specialist models happen there too, deterministically, rather than being farmed out to a service you cannot inspect. Where you choose to reach beyond the edge, you do it on your terms, as an explicit decision rather than a hidden default.

Inside that boundary sit 50 specialist brains, 25 domain and 25 operational, working under deterministic governance rather than a single opaque model guessing at everything. A domain brain handles the substance of the work. An operational brain handles the discipline around it: routing, memory, verification, and the guardrails that keep the system honest. Because this all lives on your hardware, the latency is yours to tune and the data never has to leave to get a good answer.

• Local inference on the customer's own hardware, on premises and air gapped if required, with zero data egress and no public cloud round trip.
• 50 specialist brains, 25 domain plus 25 operational, coordinated under deterministic governance instead of one black box.
• A cryptographically signed audit record on every action, the Open Audit Record, so every decision has a verifiable trail.
• Post-quantum signing with ML-DSA-65, so the audit trail holds up against tomorrow's attackers, not just today's.
• Memory the customer owns, held locally, so your context never becomes someone else's training set.

Lock in is a data problem before it is a contract problem

People tend to think of vendor lock in as a commercial trap, the renewal you cannot walk away from. That is real, but it is downstream. The deeper lock in is that your data, your context, and your institutional memory end up living inside a system you do not control. Once your history is somebody else's asset, leaving is not a negotiation, it is a migration you cannot fully complete, because the most valuable part, the accumulated context, was never yours to take.

We designed Mickai so that the memory is the customer's from the first day. It is held on the customer's hardware and it stays there. Nothing about the design assumes a permanent relationship with any operator, and Mickai is held privately by its founder, so the substrate itself is not a lever anyone can pull against you later. If you own the hardware, own the memory, and own the audit trail, then staying is a choice you keep making, not a wall you cannot climb.

“Sovereignty is not a feeling of safety. It is the property that nothing important leaves the building unless you decide it should.”

Governance you can prove, not just assert

Running AI on your own hardware solves the egress problem, but it does not by itself solve the trust problem. A system that acts on your behalf needs to be accountable for what it did and why. That is why every action produces a cryptographically signed audit record, the Open Audit Record. It is not a log that can be quietly edited after the fact. It is a signed trail, and the signatures use post-quantum cryptography, ML-DSA-65, so the record stays defensible as the threat landscape shifts.

For regulated work this changes the conversation. Instead of asking your teams to trust that the model behaved, you can show, action by action, what happened, in an order that cannot be rewritten. Determinism in the governance layer means the same inputs produce the same controlled behaviour, so an auditor is looking at a system that is meant to be inspected, not a probabilistic black box that shrugs when asked to explain itself.

The intellectual foundation under the claims

None of this is a slogan. The architecture is documented in depth across 104 filed UK patent applications, comprising 2,340 claims, each carrying a full specification, claims, and figures. These are filed applications building toward examination and grant, and together they set out how the specialist brains, the deterministic governance, the signed audit record, and the sovereign memory fit together as one coherent system rather than a bag of features.

We mention this because it is the difference between a company describing what it wishes it had built and one that has written down, in the detail a patent examiner requires, how the thing actually works. The specification is the receipt for the architecture.

Signal that the direction is landing

We are careful not to overclaim, so we will point only to signal we can stand behind publicly. On Crunchbase, our founder now ranks number 2, and the company Heat Score has reached 94 out of 100, having climbed from single digits. That is a public, third party measure of momentum, and it tells us that the appetite for sovereign AI that the customer genuinely owns is real and growing, not something we have talked ourselves into.

• Founder ranking number 2 on Crunchbase, a public and independent signal.
• Company Heat Score at 94 out of 100, up from single digits.
• Growing interest in AI that runs on the customer's own hardware, with control that stays local.

Where this goes

The next few years will separate organisations that treat AI as something they rent from those that treat it as infrastructure they hold. Rented intelligence is fine until the day the terms change, the prices move, or a breach somewhere upstream turns your context into somebody else's incident. Owned intelligence, running on your hardware, with a memory you keep and an audit trail you can prove, is a different footing entirely. It is the difference between depending on a supplier and depending on yourself.

We built Mickai for the second kind of organisation. Hybrid edge gives you the performance without the surrender. Sovereign design gives you the control without the ceiling. And because the whole thing sits on hardware you own, held under a model you own, there is no vendor waiting at the door with a renewal you cannot refuse. That, to us, is what AI without lock in was always supposed to mean, and it is where we intend to keep leading.