Research programme

The Private AI Pattern: a contribution to European sovereign AI infrastructure.

A formal, publishable pattern for production AI agents on customer-controlled infrastructure — extracted from live deployments, validated in production, contributed to the European open-internet stack through our Horizon Europe consortium.

What we are building and why it matters.

European mid-market organisations need AI that reasons over their data without surrendering sovereignty. The Private AI Pattern resolves this: governed schemas, contextual reasoning, bounded automation, self-improving knowledge, and data residency — proven in production. Our research objective is to advance this pattern toward formally verified sovereign execution, closing the gaps between data residency (guaranteed today) and full computational sovereignty (requiring trusted execution environments, customer-managed encryption, and attestable compute).

Our consortium.

Magic ID

Thessaloniki, Greece

Pattern originator and reference implementation. Two live commercial deployments across industrial manufacturing and institutional community operations.

CERTH/ITI

Thessaloniki, Greece

Centre for Research and Technology Hellas. Formal verification of the Private AI Pattern as a trusted execution boundary.

Bulgarian partner

Sofia, Bulgaria

Dissemination and market validation. Dissemination and market validation in the Bulgarian mid-market.

The Thessaloniki and Sofia Chambers of Commerce are dissemination partners, providing SME network access for pattern validation.

Current

7

TRL

Target

8–9

TRL

Technology readiness: TRL 7.

The Private AI Pattern has been validated through two years of internal production and two external paying customer deployments across structurally distinct domains. Current status: TRL 7 — system prototype demonstrated in operational environment. Target: TRL 8–9 with formal verification closing the sovereignty gaps.

Active research tracks.

DATA-02

Horizon Europe DATA-02 — Open Internet Stack.

The Private AI Pattern as a sovereign execution framework for commercial AI. Addresses the DATA-02 objective of enabling European organisations to operate AI without surrendering data to non-European infrastructure.

CL3

HORIZON-CL3 SecureAI — Civil Security.

Identity-verified agent interaction across civic and commercial contexts. The community deployment demonstrates the pattern in a civil-security-adjacent domain with deterministic access control.

What formal verification would enable.

The current pattern guarantees data residency but not computational sovereignty. Verification would close three gaps: attestable compute, customer-managed encryption, and verifiable agent boundaries. These properties would make the pattern adoptable in regulated domains beyond commerce — healthcare, legal, education, government services.

Attestable compute

Customer-managed encryption

Verifiable agent boundaries

Commercial applications: /technology and /platform. Research enquiries: research@magic-id.com.

We respect your privacy

We only use one cookie to remember your language preference. We don't use tracking cookies. No analytics, no tracking, no data collection — just a better experience for you.