Astraea: Sovereign Legal Intelligence for In-House Counsel and GCs

The privilege problem nobody wants to name

Every general counsel I speak to wants AI on their desk. They want first-pass contract review, clause extraction, matter triage, and obligation tracking done in seconds instead of hours. And almost every one of them carries the same unspoken worry: the moment a privileged document leaves the building and lands in a public-cloud model, the protection that document depended on is no longer something they control.

Legal professional privilege is not a setting you toggle. It rests on confidentiality being maintained. Once a privileged contract, a litigation strategy memo, or a regulatory response is transmitted to a third-party processor outside your walls, you have introduced a new party, a new jurisdiction, and a new set of disclosure exposures. Under the US CLOUD Act, a provider can be compelled to produce data held abroad. Under UK GDPR, privileged material is often special-category or otherwise sensitive. The honest position for most in-house teams has been to keep the AI away from the documents that matter most. That is exactly the wrong outcome.

Astraea is built to remove that trade-off. It is the legal Studio module inside Mickai, our sovereign AI operating system, and it reviews privileged contracts and matters on the firm's own infrastructure, on-prem and air-gapped, so nothing privileged ever crosses your boundary.

What Astraea actually does

Astraea is legal intelligence that runs where your documents already live. It reads contracts, surfaces and classifies clauses, flags deviations from your playbook, extracts obligations and key dates, and triages incoming matters by risk and priority. It supports diligence review, summarises long agreements, and helps counsel see the shape of an obligation set across a portfolio rather than one PDF at a time.

The point is not that another tool can read a contract. The point is where the reading happens. Astraea executes inside the firm's perimeter. No privileged text is sent to an external API. No matter file is replicated to a vendor tenant. The model comes to the data; the data never goes to the model provider. For a GC, that is the difference between an AI you can defend in front of a regulator or a court, and one you quietly cannot use on the work that counts.

Privilege survives because the boundary holds

Mickai is AI that regulated businesses own and run inside their own walls. Astraea inherits that architecture in full. It is on-prem and air-gapped by design, not as a premium add-on. There is no telemetry path that carries document content out, and no dependency on a public model endpoint to function.

Every action Astraea takes is written to the OAR, our tamper-evident, post-quantum-signed audit record. When Astraea reviews a contract, flags a clause, or proposes a redline, that event is recorded in a way that cannot be silently altered. For legal teams this matters twice over. It gives you a defensible chain of who saw what and when, and it gives you the evidence that privileged material stayed inside the boundary. If a question of disclosure or proportionality ever arises, you are not asserting confidentiality on trust. You can show it.

This is the substance behind the sovereign-AI thesis. Mickai is built and live, not a concept. It is supported by 104 filed UK patent applications carrying roughly 2,340 claims, owned by Mickai LTD with myself as inventor. These are filed rather than granted, and I am deliberate about that distinction: they establish priority and a prior-art moat around how sovereign, audited, on-prem AI is built. Astraea sits on top of that foundation rather than bolting governance on afterwards.

Why this is a regulated-firm wedge, not a feature

The market that legally cannot send data to public-cloud AI is large and growing. We estimate roughly 0.85 million UK businesses, around 15 percent, sit inside that constraint, and close to 5 million across the EU. The drivers are not going away: PRA SS2/21, UK GDPR special-category handling, the NHS DSP Toolkit, the EU AI Act's high-risk obligations, ITAR and EAR controls, NIS Regulations, and for many firms the long reach of the US CLOUD Act. The sovereign AI market is projected to grow from around USD 40 billion in 2025 to USD 148 billion by 2032.

Legal departments sit squarely inside that constraint. The work they most want to accelerate, contract and matter review, is precisely the work they are least able to expose. Astraea is designed for that exact gap. It is one of a family of Greek-named Studio modules in Mickai, alongside Nemesis for fraud and AML, Plutus for finance, Tyche for underwriting, Nomos for compliance, Panacea for clinical work, and others. Same sovereign substrate, same audit record, different domain. A GC adopting Astraea is not buying a point tool. They are standing up a legal function on infrastructure the rest of the regulated business can share.

Built and live, building to scale

This is real and operating today, and we are building to scale it across more firms and more legal workflows. As a dated, third-party momentum signal, as of June 2026 I was ranked number 4 on Crunchbase's CB Rank for people, verified live, with the Mickai company profile in the global top one to two percent. I treat that as a snapshot of attention rather than a permanent claim. What it tells me is that the sovereign-AI thesis is landing with the people who feel the privilege problem most acutely.

I want to be clear about our posture toward the broader AI ecosystem. Mickai is an ally, not an attempt to unseat anyone. Public-cloud models are extraordinary for open, non-sensitive work. Astraea exists for the documents you cannot put there. Most serious legal operations will run both: public models for the open material, a sovereign module for the privileged core. That dual-buyer reality is the design assumption, not an edge case.

Getting involved early

As Mickai scales, a pre-seed window is open to a selected group of partners who want to be involved early. This is an opportunity to back sovereign legal intelligence at the point where the architecture is proven and the market constraint is sharpening. If you are a general counsel who wants to pilot Astraea inside your own walls, or an investor who understands why privilege and sovereignty are about to become non-negotiable, I would welcome the conversation.

You can reach me directly at micky@mickai.co.uk.

Micky Irons, founder and CEO of Mickai.

FAQ

Does Astraea send privileged documents to an external AI provider? No. Astraea runs on-prem and air-gapped inside the firm's own boundary. Privileged contracts and matter files are never transmitted to a public-cloud model or vendor tenant. The model executes where the data already lives.

How does Astraea help preserve legal professional privilege? Privilege depends on confidentiality being maintained. Because Astraea keeps privileged material inside your perimeter and records every action in the tamper-evident, post-quantum-signed OAR audit record, you can demonstrate that nothing privileged crossed your boundary, rather than asserting it on trust.

Is Astraea a finished product or a concept? It is built and live as the legal Studio module inside Mickai, the sovereign AI operating system, and we are building to scale it across more firms. It shares the same sovereign substrate and audit record as the other Studio modules.

Can we still use public-cloud AI alongside Astraea? Yes. Mickai is an ally to the wider AI ecosystem, not a replacement for it. Most legal teams will run public models for open, non-sensitive work and use Astraea for the privileged core they cannot expose.

Frequently asked questions

Does Astraea send privileged documents to an external AI provider?

No. Astraea runs on-prem and air-gapped inside the firm's own boundary. Privileged contracts and matter files are never transmitted to a public-cloud model or vendor tenant. The model executes where the data already lives.

How does Astraea help preserve legal professional privilege?

Privilege depends on confidentiality being maintained. Because Astraea keeps privileged material inside your perimeter and records every action in the tamper-evident, post-quantum-signed OAR audit record, you can demonstrate that nothing privileged crossed your boundary, rather than asserting it on trust.

Is Astraea a finished product or a concept?

It is built and live as the legal Studio module inside Mickai, the sovereign AI operating system, and we are building to scale it across more firms. It shares the same sovereign substrate and audit record as the other Studio modules.

Can we still use public-cloud AI alongside Astraea?

Yes. Mickai is an ally to the wider AI ecosystem, not a replacement for it. Most legal teams will run public models for open, non-sensitive work and use Astraea for the privileged core they cannot expose.