By Micky Irons, founder of Mickai.
Why regulated banks cannot use public cloud AI
Mickai is a sovereign intelligence operating system that regulated businesses own and run inside their own walls. We built it for the part of banking that public cloud AI cannot lawfully reach: the correspondent flows, the special category customer data, the model risk that a supervisor will ask you to evidence years after the decision was made. We run entirely on your own hardware, on premises and air gapped, with zero data egress and no public cloud round trip. Nothing leaves the building, because there is no building to leave to.
That constraint is not a preference. Around 0.85 million UK businesses, roughly 15 percent, and about 5 million across the EU, legally cannot send data to public cloud AI. Banks sit at the centre of that population. The drivers are concrete and already in force: PRA model risk expectations under SS1/23, UK GDPR special category data, the EU AI Act high risk classification, the NIS Regulations, and the reach of the US CLOUD Act over any provider with a US nexus. A bank that pipes customer data, transaction histories, or underwriting logic into a shared cloud model is not adopting AI faster than its peers. It is accepting a legal and supervisory exposure that a well run board will not sign.
We remove the exposure by removing the round trip. The model runs where the data already sits, under the bank's own controls, inside its own network boundary. There is no third party processor to contract around, no cross border transfer to justify, no external endpoint to add to a threat model.
An audit record a supervisor can trust without trusting us
Banking runs on evidence. When a transaction is blocked, a customer refused, or a limit changed, the institution has to be able to show what happened, when, and on what basis, sometimes long after the people involved have moved on. We designed for that from the ledger up.
Every consequential action in Mickai is written to the Open Audit Record (OAR). Each entry is signed under post-quantum cryptography (FIPS 204 ML-DSA-65, with ML-KEM-768) and hash-chained into a tamper-evident, append-only ledger. Anyone can verify that ledger offline, for decades, without trusting the vendor and without a live connection back to us. The point is deliberate. A control that only the supplier can attest to is not a control a regulator should accept. The OAR lets a bank, its auditors, and its supervisor check the chain themselves, on their own machines, on their own timetable. We also offer OAR-as-a-Service for institutions that want the same provenance layer over systems we did not build.
Fifty specialist models, one deterministic answer
We run about fifty specialist models, twenty-five domain and twenty-five operational, with cross-model routing under a deterministic arbiter. The arbiter matters for banking more than the raw model count. A single general model that returns a different answer to the same question on two different days is not something a bank can put in front of an AML committee. Our routing is reproducible: the same inputs, under the same policy, produce the same output, and the OAR records the path that produced it.
The capabilities are organised into studios. The names are drawn from Greek myth, the functions are entirely serious. Nemesis handles fraud and AML. Plutus covers finance and FP&A. Tyche runs underwriting. Prometheus does forecasting. Iris handles customer service and Vinis handles voice. Nomos covers compliance, Astraea covers legal, and Aletheia handles audit. Pythia provides business intelligence, and Panacea serves the clinical work that sits alongside a bank's own health and insurance lines. Trust Agent is the perimeter, and the Agentic Marketing Team runs the outward facing content. Each studio draws on our own models, never on any external hosted service.
Attestation across sites, without a central server
A bank rarely runs one site. It runs many, often across jurisdictions that are not allowed to pool data. Pantheon, our post-quantum Layer 1 (currently on testnet), gives multi-node attestation across fielded units with no central server. Each site keeps its own sovereign instance and its own OAR, and the network lets those instances attest to one another's integrity without any of them shipping customer data to a shared point. Sovereignty at each node, verifiable consistency across the estate.
Filed patents and a clear commercial position
The architecture behind all of this is protected. We hold 104 filed UK patent applications, roughly 2,340 claims, across 13 invention families, owned by Mickai LTD, named inventor Mickarle Sean Junior Wagstaff-Irons. These are filed, not granted. Filing establishes priority and a prior-art moat, which is precisely what an institution wants to see behind a supplier it intends to embed in a regulated process.
Our commercial thesis has two sides. We sell sovereign AI to the regulated firms the public cloud cannot lawfully reach. We also license the patented stack to the platforms that want to reach those firms. Internal analysis maps 196 companies and 311 patent-company pairs as potential licensees, names including Microsoft, AWS, NVIDIA, Google, Adobe, and IBM. That is potential-licensee sizing, not a signed book and not an infringement claim. We are an ally to the AI majors, not their opponent. A platform that adds a sovereign, on premises layer instantly reaches a regulated market it cannot serve today, and the provenance and post-quantum controls come with it.
The market we are built for
The sovereign AI market is roughly USD 40 billion in 2025, rising to about USD 148 billion by 2032. The demand is not speculative. It is the direct consequence of supervisory expectations, data protection law, and export controls that already govern how banks handle information. We are Mickai LTD, a UK company, Companies House 17166618, with Birmingham manufacturing secured, led by founder and CEO Micky Irons. The operating system is built and live today. It runs inside the bank's own walls, produces reproducible outputs, and leaves behind an audit record the bank can defend on its own terms.
Our pre-seed round is opening soon, and we welcome inquiries from interested partners by email at micky@mickai.co.uk or on LinkedIn. The stronger point, for a bank evaluating us, is simpler. The regulated future of AI is the sovereign one, and it is already running.
FAQ
Does Mickai send any banking data to the cloud?
No. We run entirely on the customer's own hardware, on premises and air gapped, with zero data egress and no public cloud round trip. The model executes where the data already resides, inside the bank's own network boundary, so there is no cross border transfer and no third party endpoint to account for.
How can an auditor trust the record without trusting Mickai?
Every consequential action is signed under post-quantum cryptography (FIPS 204 ML-DSA-65, with ML-KEM-768) and hash-chained into a tamper-evident, append-only ledger, the Open Audit Record. Auditors and supervisors verify that ledger offline, for decades, on their own machines, with no live link back to us and no reliance on our word.
Are the patents granted?
They are filed, not granted. We hold 104 filed UK patent applications, roughly 2,340 claims, across 13 invention families, owned by Mickai LTD. Filing establishes priority and a prior-art moat, which is the relevant protection for an institution embedding the stack in a regulated process.
Written by Micky Irons, founder of Mickai. Originally published at https://mickai.co.uk/articles/banking. More from Mickai at mickai.co.uk.
Top comments (0)