Sovereign AI · Cyber · Systems thinking

Own the AI you depend on.

In a polarised world, the benefits of AI accrue to the organisations that control it. Sovereign AI is that control: the model, the harness, the knowledge, the safeguards and the regulation designed as one system, owned end to end, in Europe and beyond.

See the point of view

Approach

Systems thinking across AI and cyber

Ground

Seven working papers and a charter

Base

The Netherlands, EU

A point of view

What sovereign AI requires, argued in figures from the research material.

Working paper 1The leverage moved to the scaffolding.Five stages of engineering leverage, each one further from the model weights, and each one an organisation can own without owning the frontier.

Working paper 2Seven layers, one enclave, two crossings.The reference architecture: a request descends the enclave, its evidence ascends the log, and the boundary carries exactly two crossings.

Working paper 4Eight control points, not two classifiers.Where guardrails actually sit on the request path, and why the structural points are the ones that move bypass rates to near zero.

Working paper 7Four scenarios for 2027 to 2032.Two uncertainties, four futures, and what would falsify each. The architecture is built to survive the one Europe is in today.

Evangelist and builder

Self-sufficiency in AI is how an organisation reaps its benefits.

Kirti Kumar is an evangelist and a builder. He brings a systems-thinking mindset to AI and cyber: the model, the harness around it, the knowledge it runs on, the controls that gate it, the regulation that binds it and the economics that pay for it are one system, and they have to be designed as one.

His belief is direct. In a polarised world, European and global organisations need to create self-sufficiency in AI to reap its benefits. Not autarky, and not a rented dependency dressed up as a strategy, but a capability they control: sovereign by design, cyber-proof by construction and compliant by evidence.

That is a point of view, and it is grounded in the research material behind it: seven working papers and a charter, with forty-one figures, control catalogues and worked use cases, written and revised as the field moves. Members are invited into a dialogue on what sovereign AI actually requires, so that as an industry we move the dot forward.

Role

Evangelist · builder · systems thinker

Belief

Self-sufficiency in AI, for European and global organisations

Basis

The research material: seven working papers and a charter

Invitation

A dialogue on what sovereign AI requires

Themes

Four themes the research material argues.

Each theme is worked through to something that runs, something that can be measured, and something a regulator can read.

Sovereign AI architecture

A layered reference architecture with a sovereignty score for every layer, tiered workloads, an owned model portfolio, and a policy-gated valve to the frontier. Sized from the workload, not the brochure.

Cyber-proofing AI systems

Guardrails as control points on the request path, the harness as the enforcement point, a signed model supply chain, and detection rules that join what an agent reads, does and reaches.

Regulation as engineering

Roles, dates and evidence for the AI Act, NIS2, DORA, the CRA and the GDPR, treated as one pipeline that produces the technical file from the system's own telemetry.

AI for the security operation

A sovereign security operations model in which the agents that defend the estate are themselves governed, monitored and kept inside the enclave.