Sovereign AI for Maritime and Shipping: Sanctions, AIS-Gap and Dark-Fleet Screening On-Prem

The risk sits in the gaps, not the ledger

Sanctions exposure in shipping does not announce itself. It hides in a beneficial owner three shells deep, a flag hopped between registries in a single quarter, a ship-to-ship transfer off a coast with no port call, and above all in the AIS gap, the quiet hours when a transponder goes dark and a vessel that was in one basin reappears near a European terminal with a different draft and a cleaner story. The dark fleet moving sanctioned crude and refined product has turned voyage risk into a screening problem that ordinary compliance tooling was never built to hold.

For a marine underwriter or a shipping operator, the regulatory teeth are real. OFSI in the UK and OFAC in the US both expect you to know the vessel, the ownership chain, the cargo and the voyage pattern before you bind cover or clear a fixture. Strict-liability sanctions regimes do not care that the ownership was deliberately obscured. The G7 price cap on Russian oil pushed attestation and diligence obligations directly onto insurers, brokers and traders. A single missed match can mean a designated party on your book, a frozen claim and an enforcement action that names your firm.

So the screening has to be good. And here is the tension that has held the sector back: the data you most want to reason over is exactly the data you least want to send to someone else's cloud.

Why the cloud is the wrong place to screen a vessel

Effective dark-fleet screening means fusing voyage data, satellite AIS, ownership graphs, P&I records, cargo documents, prior claims and internal underwriting notes into one judgement. That fused picture is commercially sensitive, often personal under UK GDPR, and sometimes export-controlled. Pushing it into a public large-language-model endpoint means your live book, your intelligence sources and your risk appetite leave your control and, under the US CLOUD Act, sit within reach of a foreign production order. For a regulated carrier answering to the PRA and the FCA on operational resilience, that is not a trade you can make.

Mickai removes the trade entirely. Mickai is a sovereign AI operating system, an SIOS: AI that a regulated business owns and runs inside its own walls, on-prem and air-gapped. The models, the vessel and ownership corpus, the sanctions lists and the reasoning never leave the building. There is no external endpoint to leak to and no third-party log to subpoena. It is built and LIVE, and we are building to scale.

Screening a voyage without the data ever leaving

Inside Mickai the work runs through the Studios, our Greek-named application suites, coordinated by a deterministic arbiter over fifty specialist brains. For maritime sanctions and dark-fleet work the relevant Studios lock together:

• **Nemesis** handles fraud and AML: it walks the ownership graph through shell layers, flags registries and managers with prior designations, and scores a vessel against the pattern of known evasion behaviour rather than a static list.
• **Nomos** carries the compliance logic: OFSI and OFAC consolidated lists, the G7 price-cap attestation rules, NIS Regs obligations and your own internal policy, applied consistently to every case.
• **Tyche** brings the underwriting view, turning the fused signal into a bind or decline recommendation with the voyage-risk reasoning attached.
• **Aletheia** is the audit brain, writing every step down.

The AIS gap is treated as a first-class signal, not a footnote. When a transponder goes dark, Mickai reasons over the last known position, the plausible range, the reappearance point and the change in reported draft, and surfaces the ship-to-ship transfer that the operator hoped no one would reconstruct. Because the retrieval layer is an air-gapped RAG built over your own vessel and ownership corpus, the model answers from your intelligence, not from stale public training data, and it never phones home to do it.

The record that survives an enforcement review

Screening a vessel is only half the obligation. The other half is proving, months later, that you screened it properly, on the facts you held at the time. Mickai writes every action to the OAR, an Objective Audit Record: a tamper-evident log signed with ML-DSA-65, a post-quantum signature standard, so the trail cannot be altered after the fact and stays verifiable even against future cryptographic attack. Identity is hardware-bound, so each decision ties to a specific analyst on a specific machine. Where a screening decision needs to be unwound, compensating rollback reverses it cleanly without erasing the history that a regulator will ask to see.

For the buyer roles that own this exposure, the OAR is the point. The MLRO gets a defensible sanctions-screening trail. The Chief Risk Officer gets voyage risk quantified and evidenced. The General Counsel gets a record that holds up under an OFSI or OFAC review. The CISO gets an architecture where the crown-jewel data never crosses the perimeter. And the Board gets operational resilience it can attest to under the PRA and FCA regime without hoping a cloud vendor behaves.

Why this is defensible

Mickai is protected by 104 filed UK patent applications, roughly 2,340 claims, held by Mickai LTD with inventor Micky Irons. Filed, not granted: what that gives us is priority and a prior-art moat across the sovereign architecture, the air-gapped RAG, the post-quantum-signed audit record and the deterministic arbiter over fifty brains.

The market is moving with us. Sovereign AI was valued around USD 40 billion in 2025 and is projected toward USD 148 billion by 2032, and roughly 0.85 million UK businesses and around 5 million across the EU are legally barred from putting sensitive workloads into public-cloud AI at all. As a dated third-party momentum signal, in June 2026 Micky Irons was ranked number four on Crunchbase, with the Mickai company profile in the top one to two percent globally. We are a UK company with Birmingham manufacturing secured, positioned as an ally to the wider AI ecosystem rather than a challenger to any single lab.

The window

We are opening a pre-seed window to a small number of selected partners, insurers, shipping principals and infrastructure backers who understand that sovereign AI is where regulated data has to live. This is a deliberate, selective process. If maritime sanctions and dark-fleet exposure sit on your book, and you want to see vessel and voyage screening run entirely inside your own walls, I would welcome the conversation.

Micky Irons, founder and CEO of Mickai. micky@mickai.co.uk

FAQ

Can Mickai screen vessels and ownership without sending data to a public cloud? Yes. Mickai runs on-prem and air-gapped inside your own environment. The models, the vessel and ownership corpus, the sanctions lists and the reasoning stay behind your perimeter, so nothing is exposed to an external endpoint or reachable under the US CLOUD Act.

How does Mickai handle the AIS gap and dark-fleet behaviour? The AIS gap is treated as a primary signal. Mickai reasons over the last known position, the plausible range during the dark period, the reappearance point and any change in reported draft to reconstruct probable ship-to-ship transfers and evasion patterns, then scores the voyage against known dark-fleet behaviour rather than a static list alone.

How does Mickai help me prove OFSI and OFAC compliance after the fact? Every screening action is written to the OAR, a tamper-evident audit record signed with the post-quantum ML-DSA-65 standard and tied to a hardware-bound identity. That gives your MLRO and General Counsel a verifiable trail of exactly what was screened, on what facts, by whom, that holds up under an enforcement review.

Is Mickai a live product or a roadmap? Mickai is built and LIVE, and we are building to scale. It is a sovereign AI operating system protected by 104 filed UK patent applications covering the air-gapped architecture, the post-quantum-signed audit record and the deterministic arbiter over fifty brains.

Frequently asked questions

Can Mickai screen vessels and ownership without sending data to a public cloud?

Yes. Mickai runs on-prem and air-gapped inside your own environment. The models, the vessel and ownership corpus, the sanctions lists and the reasoning stay behind your perimeter, so nothing is exposed to an external endpoint or reachable under the US CLOUD Act.

How does Mickai handle the AIS gap and dark-fleet behaviour?

The AIS gap is treated as a primary signal. Mickai reasons over the last known position, the plausible range during the dark period, the reappearance point and any change in reported draft to reconstruct probable ship-to-ship transfers and evasion patterns, then scores the voyage against known dark-fleet behaviour rather than a static list alone.

How does Mickai help me prove OFSI and OFAC compliance after the fact?

Every screening action is written to the OAR, a tamper-evident audit record signed with the post-quantum ML-DSA-65 standard and tied to a hardware-bound identity. That gives your MLRO and General Counsel a verifiable trail of exactly what was screened, on what facts, by whom, that holds up under an enforcement review.

Is Mickai a live product or a roadmap?

Mickai is built and LIVE, and we are building to scale. It is a sovereign AI operating system protected by 104 filed UK patent applications covering the air-gapped architecture, the post-quantum-signed audit record and the deterministic arbiter over fifty brains.