Sovereign AI for Water Utilities: Operational Intelligence Without Sending SCADA to the Cloud

The problem the control room cannot ignore

A water utility is critical national infrastructure. The data that runs it lives in SCADA systems, telemetry historians, pump and valve controllers, water-quality sensors, and the operational technology that keeps clean water flowing and wastewater treated. None of it was designed to leave the plant. Under the Network and Information Systems (NIS) Regulations, operators of essential services carry a legal duty to manage the security of those systems and to demonstrate, to a competent authority, that they have done so.

That duty collides with the way modern AI is usually sold. The dominant pattern is a cloud endpoint: stream your data out, get an answer back. For a back-office team that may be a calculated trade. For a water operator pushing real-time SCADA telemetry through a third party's infrastructure, it is a non-starter. The moment operational data crosses the boundary, the utility has expanded its attack surface, weakened its segmentation story, and handed a regulator a question it does not want to answer.

So most operators do the rational thing. They keep AI out of the operational environment entirely. The intelligence stays in slide decks and quarterly reviews while the control room runs on alarms and hard-won experience. That is the gap Mickai was built to close.

Why air-gapped is the only honest answer for OT

The instinct to keep SCADA off the cloud is not timidity. It is correct engineering. Operational technology sits behind layers of network segmentation for a reason, and the reference models the sector relies on treat any path from the control network to the public internet as a liability to be eliminated, not managed. Bolting a cloud AI dependency onto that environment reverses years of careful isolation work.

Mickai takes the opposite stance. It is a sovereign AI operating system, or SIOS: AI that a regulated business owns and runs entirely inside its own walls, on-premises and air-gapped. The model weights, the inference, the retrieval, the orchestration, all of it executes on hardware the operator controls. No telemetry leaves the plant. No operational data is sent to a vendor for processing, training, or any other purpose. The AI is brought to the data, behind the existing segmentation, rather than the data being pumped out to the AI.

This is the difference between an assistant you can put in front of a NIS competent authority and one you have to keep out of the building. With Mickai, the operational network stays closed. The intelligence still arrives.

What operational intelligence looks like over real plant data

Once the boundary problem is solved, the use cases that were previously off-limits open up. Across the Mickai Studios, several map directly onto water operations:

• **Prometheus** handles forecasting: demand modelling, flow and pressure prediction, and early warning on conditions that precede a burst main or a treatment excursion. Forecasting over your own historian, not a generic model that never saw your network.
• **Aletheia** is the audit Studio. Every inference, every retrieval, every operator query is captured as evidence, which matters when you have to show a regulator how a decision was reached.
• **Nomos** carries compliance reasoning, mapping operational events against the obligations that apply to an operator of essential services so that incident handling and reporting are grounded in the actual rules.
• **Pythia** delivers business intelligence over operational and asset data without exporting any of it.

These run as part of one owned system rather than a sprawl of disconnected cloud subscriptions. The operator gets forecasting, anomaly insight, compliance support, and analytics over live plant data, and none of it depends on a connection the security team spent a decade closing.

The audit trail a regulator can actually verify

Sovereignty is not only about where data sits. It is about being able to prove what the system did. Every action Mickai takes is written to the Operational Audit Record, the OAR: a tamper-evident log, post-quantum-signed, that records what the AI was asked, what it retrieved, and what it returned.

For a water operator, this changes the conversation with a competent authority from assertion to evidence. Instead of describing a control, you produce a signed, ordered record of every interaction the AI had with operational data. If a decision is questioned months later, the record stands on its own. Post-quantum signing means that evidentiary value is built to survive the cryptographic transitions coming over the asset lifetimes these utilities plan around. An audit trail that can be quietly edited is not an audit trail. The OAR is designed so it cannot be.

Why this is a category, not a feature

The water sector is one instance of a much larger structural pattern. By our analysis, roughly 0.85 million UK businesses, about fifteen percent, and around five million across the EU are legally barred from putting their most sensitive workloads on public-cloud AI. The drivers are the regulatory frameworks that critical-infrastructure and regulated operators already live inside: the NIS Regulations, UK GDPR special-category rules, the EU AI Act high-risk obligations, the extraterritorial reach of the CLOUD Act, and sector regimes beyond. The sovereign AI market is sized at around USD 40 billion in 2025, on a path toward roughly USD 148 billion by 2032. Water utilities are not a niche. They are an early, clear-cut example of a buyer who cannot use the cloud answer and has been waiting for a real one.

That position is underwritten by a deep IP estate: 104 filed UK patent applications, some 2,340 claims, held by Mickai LTD, with myself as inventor. Filed, not granted, which establishes a priority position and a prior-art moat over the architecture of running owned, audited AI inside the walls of a regulated operator. As a third-party momentum signal, in June 2026 I was ranked number four on Crunchbase, with Mickai placing in the top one to two percent of companies globally. The system is built and live, and we are building to scale, with UK manufacturing secured in Birmingham.

A note on posture: Mickai is an ally to the broader AI ecosystem, not a challenger to it. The hyperscalers serve the workloads that belong in the cloud. We serve the ones that legally cannot go there. That is a complementary category, and it is a category a hyperscaler would rationally want to own.

Working with water-sector partners

We are working with a small number of selected partners, water operators and critical-infrastructure operators who want to deploy operational intelligence over their own SCADA and operational data without compromising their NIS position. This is selective by design rather than a broad campaign. The goal is a handful of reference deployments where the air-gapped architecture and the OAR audit trail are proven in a live operational setting.

If you operate essential water services and the gap between what AI could do and what your security model allows has been frustrating you, this is the conversation to have. Reach me directly at micky@mickai.co.uk.

Micky Irons, founder and CEO of Mickai.

FAQ

Does Mickai send any SCADA or operational data to the cloud? No. Mickai is a sovereign AI operating system that runs on-premises and air-gapped. The model weights, inference, retrieval, and orchestration all execute on hardware the operator controls. No telemetry or operational data leaves the plant for processing, training, or any other purpose.

How does Mickai help with NIS Regulations compliance? The AI stays behind the operator's existing network segmentation, so adopting it does not add a path from the control network to the public internet. Every action is written to the Operational Audit Record, a tamper-evident, post-quantum-signed log, giving a competent authority verifiable evidence rather than assertions.

What is the OAR? The Operational Audit Record is a tamper-evident, post-quantum-signed log of what the AI was asked, what it retrieved, and what it returned. It is designed so the record cannot be quietly edited, preserving evidentiary value across the long asset lifetimes water utilities plan around.

Which Mickai Studios apply to water operations? Prometheus for forecasting (demand, flow, pressure, early warning), Aletheia for audit, Nomos for compliance reasoning, and Pythia for business intelligence over operational and asset data. They run as one owned system rather than separate cloud subscriptions.

Is this only relevant to water utilities? No. Water is one clear example of a broader category. Roughly 0.85 million UK businesses and around five million across the EU are legally barred from public-cloud AI under regimes such as the NIS Regulations, UK GDPR, the EU AI Act, and the CLOUD Act. The sovereign AI market is sized at around USD 40 billion in 2025, on a path toward roughly USD 148 billion by 2032.

Frequently asked questions

Does Mickai send any SCADA or operational data to the cloud?

No. Mickai is a sovereign AI operating system that runs on-premises and air-gapped. The model weights, inference, retrieval, and orchestration all execute on hardware the operator controls. No telemetry or operational data leaves the plant for processing, training, or any other purpose.

How does Mickai help with NIS Regulations compliance?

The AI stays behind the operator's existing network segmentation, so adopting it does not add a path from the control network to the public internet. Every action is written to the Operational Audit Record, a tamper-evident, post-quantum-signed log, giving a competent authority verifiable evidence rather than assertions.

What is the OAR?

The Operational Audit Record is a tamper-evident, post-quantum-signed log of what the AI was asked, what it retrieved, and what it returned. It is designed so the record cannot be quietly edited, preserving evidentiary value across the long asset lifetimes water utilities plan around.

Which Mickai Studios apply to water operations?

Prometheus for forecasting, Aletheia for audit, Nomos for compliance reasoning, and Pythia for business intelligence over operational and asset data. They run as one owned system rather than separate cloud subscriptions.

Is this only relevant to water utilities?

No. Water is one clear example of a broader category. Roughly 0.85 million UK businesses and around five million across the EU are legally barred from public-cloud AI under regimes such as the NIS Regulations, UK GDPR, the EU AI Act, and the CLOUD Act. The sovereign AI market is sized at around USD 40 billion in 2025, on a path toward roughly USD 148 billion by 2032.