A 104-patent moat in sovereign AI

Why intellectual property is the real moat in sovereign AI

In artificial intelligence, the model is rarely the durable advantage. Weights get cheaper, sovereign models improve every quarter, and capability gaps close fast. What does not close fast is the legal ground underneath a category. That ground is intellectual property, and in sovereign AI it is where Mickai has spent its earliest and most deliberate capital.

Mickai is the sovereign AI operating system. It is AI that regulated businesses own and run inside their own walls, on-premise and air-gapped, with every action written to a tamper-evident, post-quantum-signed audit record we call the OAR. This is built and live, not a concept. The portfolio behind it stands at 104 filed UK patent applications, 2,340 claims, owned by Mickai LTD with myself as inventor.

These are filed, not granted. That distinction matters, and I will be precise about it. Filed applications establish priority and create a prior-art moat from the filing date forward. They do not yet confer the enforceable rights that come with grant. We say filed because that is the honest, defensible status, and because diligence-grade buyers respect precision far more than inflated claims.

What the 104 applications actually cover

The portfolio maps to the architecture of how sovereign AI is owned, governed, and trusted.

The core covers the operating-system boundary itself: how an AI runtime sits alongside Windows and Linux for AI activity only, sandboxed, isolated, and gated from the host. It covers the tamper-evident, post-quantum-signed audit record, the mechanism that lets a regulated firm prove after the fact exactly what its AI did, when, and on whose authority. As adversaries move toward quantum-capable attacks, signing that record with post-quantum cryptography is not a flourish. It is the difference between an audit trail that survives the next decade and one that does not.

Around that core sit applications covering memory and retrieval that stay inside the sovereign boundary, drift and hallucination controls, and the Studio modules that turn the substrate into working software for regulated functions: Nemesis for fraud and AML, Plutus for finance and FP&A, Tyche for underwriting, Prometheus for forecasting, Iris for customer service, Nomos for compliance, Astraea for legal, Panacea for clinical, Pythia for business intelligence, and Aletheia for audit. Alongside them sit Trust Agent, the AMT agentic marketing layer, the Vinis voice system, and OAR-as-a-Service. The portfolio reads as a map of the category, not a scattering of opportunistic filings.

The wedge the moat protects

A moat is only worth building around something people are required to buy. Sovereign AI has that demand structure set by regulation, not by hype.

There is a large class of organisations that legally cannot send their data to public-cloud AI. Around 0.85 million UK businesses, roughly 15 percent, sit in that position, and approximately 5 million across the EU. The drivers are concrete: PRA SS2/21 on model risk, UK GDPR special-category handling, the NHS DSP Toolkit, the EU AI Act high-risk regime, ITAR and EAR export controls, the NIS Regulations, and the extraterritorial reach of the US CLOUD Act. These firms do not want sovereign AI as a preference. Their compliance teams require it. The sovereign AI market is sized at around USD 40 billion in 2025, growing toward USD 148 billion by 2032. The 104 applications stake a claim across the architecture serving exactly that demand.

The licensing thesis

This is where the portfolio stops being purely defensive and becomes an asset on its own.

Our analysis maps 196 companies and 311 patent-company pairs as potential licensees, including Microsoft, AWS, NVIDIA, Google, Adobe, and IBM. I want to be exact about what that is and is not. This is potential-licensee sizing, an assessment of where our filed claims read onto the direction these companies are already moving in sovereign, on-premise, and auditable AI. It is not an assertion that anyone is infringing anything. Each of those names is building toward regulated, sovereign deployment, and foundational IP at the OS-boundary and audit-record layer is the kind of position large platforms typically license rather than litigate around.

That is also why I frame Mickai as an ally, not an OpenAI killer. The dual-buyer thesis is that regulated enterprises buy the sovereign operating system to run their own AI, and the largest platform companies are natural licensees of the underlying IP. Both sides of that can be true at once, and the patent portfolio is the connective tissue.

Acquirer-grade defensibility

When a strategic acquirer or a serious investor runs diligence on an AI company, the model weights are rarely the prize, because anyone can train weights. What survives diligence is defensible, original IP that establishes priority in a category with regulatory tailwinds. A 104-application portfolio with 2,340 claims, owned cleanly by a single UK entity with a single named inventor, is exactly the structure that holds up under that scrutiny. Clean ownership, clear inventorship, early priority dates, and claims that track a real and growing market.

As a dated, third-party momentum signal, in June 2026 I was ranked number 4 on Crunchbase by CB Rank for people, verified live, with the Mickai company entity in the top 1 to 2 percent globally. That is a snapshot from June 2026, not a permanent claim, and I treat it as one external read among many. The substance is the portfolio, the live product, and the regulated demand it serves. Mickai is a UK company with Birmingham manufacturing secured, and we are building to scale.

The honest status, and what comes next

I will restate the most important caveat plainly, because it is the one that earns trust. These 104 applications are filed, not granted. Filing buys priority and a prior-art position from the filing date. Grant is a process that takes time and examination, and we will report it accurately at every stage. Anyone who tells you a filed application is the same as an enforceable granted patent is not someone you want drafting your diligence memo.

A window for selected partners

We are speaking with a small number of selected partners as we scale. This is an invitation to be early in a sovereign AI operating system that is already built and live, backed by one of the most deliberate IP positions in the category. It is an opportunity. The product runs, the demand is regulatory, and the moat is filed.

If you want the detail behind the portfolio, the licensing analysis, or the wedge, reach me directly at micky@mickai.co.uk.

By Micky Irons, founder and CEO of Mickai.

Frequently asked questions

Are Mickai's 104 patents granted?

No. They are 104 filed UK patent applications, 2,340 claims, owned by Mickai LTD with Micky Irons as inventor. Filed, not granted. Filing establishes priority and a prior-art position from the filing date, but it is not the same as an enforceable granted patent. We report status accurately at every stage.

What do the patents actually cover?

The architecture of sovereign AI: the operating-system boundary that runs AI alongside Windows and Linux in a sandboxed, gated way; the tamper-evident, post-quantum-signed audit record (OAR); memory and retrieval that stay inside the sovereign boundary; drift and hallucination controls; and the Greek-named Studio modules for regulated functions such as fraud and AML, finance, underwriting, compliance, legal, clinical, and audit.

Who is forced to buy sovereign AI?

Regulated organisations that legally cannot send data to public-cloud AI. Around 0.85 million UK businesses, roughly 15 percent, and approximately 5 million across the EU. The drivers include PRA SS2/21, UK GDPR special-category rules, the NHS DSP Toolkit, the EU AI Act high-risk regime, ITAR and EAR, the NIS Regulations, and the US CLOUD Act. The sovereign AI market is sized at around USD 40 billion in 2025, growing toward USD 148 billion by 2032.

What is the licensing thesis?

Our analysis maps 196 companies and 311 patent-company pairs as potential licensees, including Microsoft, AWS, NVIDIA, Google, Adobe, and IBM. This is potential-licensee sizing based on where our filed claims read onto the direction these companies are already moving, not an assertion that anyone is infringing. Foundational IP at the OS-boundary and audit-record layer is the kind of position large platforms tend to license.

What is the Crunchbase ranking?

As a dated third-party momentum signal verified live in June 2026, Micky Irons ranked number 4 on Crunchbase by CB Rank for people, with the Mickai company entity in the top 1 to 2 percent globally. It is a June 2026 snapshot, not a permanent claim, and one external read among many.

How can I get involved?

We are speaking with a small number of selected partners as we scale. It is an opportunity to be early. For the detail behind the portfolio, the licensing analysis, or the wedge, contact Micky Irons directly at micky@mickai.co.uk.