Model Independence for the Enterprise

The quiet dependency almost nobody planned for

Over the last few years the enterprise quietly outsourced its thinking. Contracts, medical notes, board papers, source code, and customer records now travel to a small cluster of foreign cloud models, get processed on infrastructure the customer cannot see, and come back as answers nobody can fully trace. It happened fast, it happened without a formal decision in most organisations, and it created a dependency that would be very expensive to unwind if the terms ever changed.

We built Mickai, our Sovereign Intelligence Operating System, because we do not think that arrangement is safe for serious institutions. A hospital, a bank, a defence supplier, or a law firm should not have to send its most sensitive material across a border and hope for the best. It should be able to run capable intelligence on hardware it owns, under rules it sets, with a record it can prove. That is what model independence means to us, and it is the whole point of the system.

What dependency actually costs

The convenience of a cloud model is real, and we understand why people reached for it. The cost is less visible, so it rarely shows up in the business case until something goes wrong. It is worth naming plainly.

• Data leaves the building. Every prompt is a transfer of information to infrastructure you do not operate, governed by terms you did not write.
• The model can change underneath you. A provider can retire a version, alter its behaviour, or shift its policies, and your workflows change with it whether you like it or not.
• Pricing and access are somebody else's decision. The rate you pay and the capacity you get are set outside your organisation and can move without warning.
• You cannot inspect what you rely on. When a regulator, an auditor, or a court asks how a decision was reached, pointing at a remote black box is not an answer.
• Jurisdiction follows the data. Where your information is processed determines whose laws apply to it, and that is rarely the jurisdiction you would have chosen.

None of these are hypothetical edge cases. They are the ordinary conditions of renting intelligence from a handful of providers on the other side of an ocean. Independence is not about hostility to any of them. It is about not building your most important operations on a foundation you neither own nor can examine.

A controllable local system instead

The alternative we offer is straightforward to state and harder to build, which is why we spent years building it. Mickai runs on the customer's own hardware, on premises and air gapped when required, with zero data egress and no public cloud round trip. The sensitive material never leaves the room. The intelligence comes to the data rather than the data being shipped to the intelligence.

Inside that boundary sit 50 specialist brains, 25 domain and 25 operational, working under deterministic governance rather than a single opaque model doing everything at once. That structure matters for independence. It means capability is composed from parts the customer can reason about, each with a defined job, coordinated by rules the customer can set, rather than a monolith whose behaviour can only be observed from the outside. When the work is decomposed and governed, it becomes something an institution can actually stand behind.

Every action the system takes is written to a cryptographically signed audit record, the Open Audit Record, so there is a tamper evident account of what happened and why. We sign it with post-quantum cryptography, ML-DSA-65, because a record that has to hold up years from now should not be secured by a signature that a future machine could forge. And the memory the system accumulates belongs to the customer. It is not a training asset for someone else, and it does not walk out of the building with a vendor relationship.

“Independence is not a slogan. It is a property you can test. Can you unplug the network cable and still work? Can you prove, cryptographically, what the system did? Can you keep your own memory? If the answer is yes, you are not dependent. If it is no, you are renting.”

Why sovereignty and capability are not a trade

The old objection to running things locally was that you had to give up quality to get control. We do not accept that framing, because the assumption behind it is that only a giant remote model can be useful. In practice, most enterprise work is not open ended world knowledge. It is bounded, repetitive, and specific to the organisation: reconciling records, drafting within a house style, checking compliance against known rules, triaging cases, reasoning over documents the enterprise already holds. That is exactly the shape of work a governed set of specialist brains, sitting close to the customer's own memory, is built to do well.

Running locally also removes a whole category of friction. There is no latency to a distant data centre, no rate limit imposed from outside, no surprise deprecation, and no invoice that scales with every query. The economics stop being a meter that runs faster the more useful the system becomes. For an institution that wants intelligence woven through everyday operations rather than rationed by cost, that shift is the difference between an experiment and an operating capability.

The moat underneath the claim

We do not ask anyone to take the architecture on faith. The methods behind Mickai are set out in 104 filed UK patent applications, roughly 2,340 claims in total, each with a full specification, detailed claims, and figures, building toward examination and grant. That body of work covers how the specialist brains are governed, how the audit record is produced and signed, and how a sovereign system holds together as one coherent whole rather than a bag of parts. It is the written record of a system designed to be independent from the ground up, not retrofitted for it.

The market is beginning to notice. Our founder now ranks number 2 on Crunchbase, and the company Heat Score reached 94 out of 100, climbing from single digits. We mention that not as a victory lap but as a signal: the appetite for intelligence an institution can actually own is not niche, and it is growing quickly as the risks of the rented model become clearer to the people who carry the liability.

Where this goes next

We think the next few years will make model independence a board level requirement rather than a technical preference. Regulators will keep asking harder questions about where data is processed and how decisions are made. Boards will keep asking what happens if a provider's terms, prices, or politics shift. And the organisations that thought ahead will already have their most sensitive intelligence running inside their own walls, on their own hardware, under their own governance, with a signed record of everything it did.

That is the future we are building toward, and it is why Mickai exists in the form it does. Capable intelligence should belong to the institution that depends on it. The data should stay home, the memory should be owned, the record should be provable, and the whole thing should keep working even when the network does not. Independence is not a feature we bolted on. It is the reason we started.