Tyche for Parametric and Cyber Underwriting: Air-Gapped Risk Pricing Insurers Can Defend

The lines that move faster than the model can be governed

Parametric and cyber are the two underwriting lines where the world changes faster than a control framework can keep up. A new ransomware strain, a shifting flood map, a satellite weather feed, a fresh CVE in widely deployed software: each one re-prices risk in days, not quarters. Underwriters need AI that can ingest signals and price exposure at that speed. The same speed is exactly what makes regulators, reinsurers, and boards nervous. If a model can re-price a book overnight, who can show why it did, and prove the reasoning was not tampered with after the loss?

That is the question Tyche is built to answer. Tyche is the underwriting Studio inside Mickai, the sovereign AI operating system: AI that a regulated carrier owns and runs inside its own walls, on-prem and air-gapped, with every action written to a tamper-evident, post-quantum-signed audit record. Tyche prices emerging parametric and cyber risk where the data lives, and it produces an evidence trail the carrier, the regulator, and the reinsurer can all read the same way.

Why fast lines need on-prem AI specifically

Public-cloud AI forces a carrier to send loss data, claims history, and exposure schedules to infrastructure it does not control. For mainstream lines that is a procurement debate. For cyber and parametric it is a structural problem. Cyber underwriting data is itself a threat-intelligence asset: aggregated, it maps which insureds run which vulnerable systems. Parametric programs often touch critical-infrastructure, sovereign, or sanctioned-adjacent exposures that carry their own data-residency and export constraints.

The compliance perimeter is not abstract. PRA SS2/21 sets expectations for model risk management that a carrier must be able to evidence. UK GDPR special-category rules, the EU AI Act's high-risk classification for certain underwriting uses, NIS Regulations, ITAR and EAR for defence-adjacent exposures, and the CLOUD Act's reach over data held by US providers all bear on where and how a pricing model may run. Around 0.85 million UK businesses, roughly 15 percent of the economy, and close to 5 million across the EU are effectively barred from running this class of workload on public-cloud AI. The sovereign AI market is sized at around USD 40 billion in 2025 and on a path to USD 148 billion by 2032. Tyche is built for the carriers inside that perimeter, not the ones who can offload the problem to someone else's data centre.

What Tyche actually does

Tyche runs the underwriting loop end to end inside the carrier's own environment. It ingests parametric triggers and indices, cyber threat feeds, exposure schedules, and historical loss data without any of it leaving the building. It prices risk, proposes terms, flags accumulation and correlation across a book, and surfaces the drivers behind every number it produces. An underwriter sees not just a price but the features, weights, and reference data that produced it.

The difference is the audit layer. Every model run, every input version, every parameter, and every decision is written to the Mickai audit record, signed with post-quantum cryptography so the trail cannot be forged or quietly edited after a loss event. When a parametric trigger fires or a cyber claim lands, the carrier can reconstruct exactly what the model knew, when it knew it, and how it priced the risk at bind. That is the artifact a regulator wants for SS2/21 model governance, and it is the same artifact a reinsurer wants before it accepts a cession.

The reinsurer and the regulator read the same record

Underwriting is a chain of trust. The primary carrier prices, the reinsurer takes a share, and the regulator supervises both. Today each link rebuilds its own view of the model from partial documentation and after-the-fact reports. Tyche collapses that into one inspectable record. The reinsurer can audit the pricing logic behind a treaty without the carrier exposing its raw book to a third-party cloud. The regulator can examine model behaviour without a six-week document request. The carrier defends its numbers with cryptographic evidence rather than recollection.

This is what defensible pricing means in practice. Not a model that is merely accurate, but a model whose every decision is reconstructable and provable to every party with a stake in the loss. For lines that re-price weekly, that auditability is the only thing that makes the speed safe.

Part of a sovereign estate, not a point tool

Tyche does not stand alone. It sits beside Nemesis for fraud and AML, Plutus for finance, Nomos and Astraea for compliance and legal, Prometheus for forecasting, and Aletheia for audit, all on the same air-gapped substrate and the same signed audit record. A carrier underwriting cyber with Tyche can route a suspicious submission to Nemesis, check a treaty against regulatory text with Nomos, and forecast accumulation with Prometheus, with one governance perimeter and one evidence trail across all of it.

Underpinning the platform is an IP estate of 104 filed UK patent applications, roughly 2,340 claims, held by Mickai LTD with Micky Irons as inventor. Filed, not granted, which establishes priority and a prior-art position across sovereign on-prem AI and tamper-evident audit. The same architecture maps to 196 companies and 311 patent-company pairs as potential licensees, including the largest cloud and AI providers. That is potential-licensee sizing, not committed revenue, but it frames why this is infrastructure a hyperscaler would rather own than build around.

The momentum is visible externally. In June 2026, Micky Irons was independently ranked number 4 on Crunchbase, with Mickai placing in the top 1 to 2 percent of companies globally. The platform is built and live, manufacturing is secured in Birmingham, and the build is scaling toward the full Studio estate.

The dual-buyer thesis

Tyche serves two buyers from one architecture. The carrier gets defensible, sovereign underwriting AI for its fastest, riskiest lines. The strategic partner sees a category: regulated AI with a built-in, cryptographically provable audit layer, sized to a market heading toward USD 148 billion, with an IP moat already filed. Mickai is built to be an ally to the wider AI ecosystem, not a replacement for it. The carriers who cannot use public-cloud AI are a market that ecosystem cannot currently serve. Tyche serves it. That is the kind of regulated-AI infrastructure a hyperscaler would want to own rather than rebuild.

Working with selected partners

I am working with a small number of selected carriers, reinsurers, and platform partners who want to price emerging risk inside their own walls with an audit trail they can defend. The aim is the right early partners in parametric and cyber, where the need for defensible on-prem pricing is sharpest and the proof points compound fastest.

If you underwrite fast-moving risk and need AI the regulator and the reinsurer can both inspect, reach me directly at micky@mickai.co.uk.

Micky Irons, founder and CEO of Mickai.

FAQ

What is Tyche? Tyche is the underwriting Studio inside Mickai, the sovereign AI operating system. It prices emerging parametric and cyber risk on-prem and air-gapped inside the carrier, and writes every model decision to a tamper-evident, post-quantum-signed audit record that the carrier, regulator, and reinsurer can all inspect.

Why does parametric and cyber underwriting need air-gapped AI? These lines re-price in days, and their data is sensitive: cyber underwriting data maps insured vulnerabilities, and parametric programs often touch critical-infrastructure and export-controlled exposures. Sending that to public-cloud AI breaches data-residency, model-risk, and export rules under frameworks like PRA SS2/21, UK GDPR, the EU AI Act, NIS Regulations, ITAR and EAR, and the CLOUD Act. Running it on-prem keeps the data and the model inside the carrier's control.

How does Tyche make pricing defensible? Every model run, input version, parameter, and decision is written to the Mickai audit record and signed with post-quantum cryptography, so the trail cannot be forged or edited after a loss. The carrier can reconstruct exactly what the model knew and how it priced at bind, giving the regulator model-governance evidence and the reinsurer audit assurance from the same record.

Is Mickai built and available now? Yes. Mickai and its Studios, including Tyche, are built and live, with manufacturing secured in Birmingham. The estate is scaling toward the full set of Greek-named Studios on the same sovereign substrate and the same signed audit record.

Frequently asked questions

What is Tyche?

Tyche is the underwriting Studio inside Mickai, the sovereign AI operating system. It prices emerging parametric and cyber risk on-prem and air-gapped inside the carrier, and writes every model decision to a tamper-evident, post-quantum-signed audit record that the carrier, regulator, and reinsurer can all inspect.

Why does parametric and cyber underwriting need air-gapped AI?

These lines re-price in days, and their data is sensitive: cyber underwriting data maps insured vulnerabilities, and parametric programs often touch critical-infrastructure and export-controlled exposures. Sending that to public-cloud AI breaches data-residency, model-risk, and export rules under frameworks like PRA SS2/21, UK GDPR, the EU AI Act, NIS Regulations, ITAR and EAR, and the CLOUD Act. Running it on-prem keeps the data and the model inside the carrier's control.

How does Tyche make pricing defensible?

Every model run, input version, parameter, and decision is written to the Mickai audit record and signed with post-quantum cryptography, so the trail cannot be forged or edited after a loss. The carrier can reconstruct exactly what the model knew and how it priced at bind, giving the regulator model-governance evidence and the reinsurer audit assurance from the same record.

Is Mickai built and available now?

Yes. Mickai and its Studios, including Tyche, are built and live, with manufacturing secured in Birmingham. The estate is scaling toward the full set of Greek-named Studios on the same sovereign substrate and the same signed audit record.