Digital and Technology Policy: National Sovereignty

Baroness Shields Excerpts

Thursday 23rd July 2026 Lords Chamber

[Baroness Shields (Con)]

My Lords, I too begin by thanking the noble Baroness, Lady Kidron, for once again bringing us together for this important conversation at a moment in time when AI is emerging as the operating system of all our lives.

In May, I cautioned this House that frontier AI systems were shaping our world, that we were being controlled by an extraordinarily narrow concentration of power, capital and data, and that most nations were not participating in that process but inheriting it. I return to that argument today because it is no longer a forecast. It is now a fact pattern. Frontier AI capability sits overwhelmingly in the hands of a few companies headquartered mainly in the United States and China. Every other nation participates by permission, not by right. That permission can be withdrawn arbitrarily in an instant.

We saw this explicitly on 12 June, when the US Department of Commerce directive forced Anthropic to suspend access to its most advanced model, Claude Fable, for foreign nationals worldwide, including its own foreign national researchers. They were unable to touch the model they had built because of the passport they held. Access was restored on 1 July. However, sovereignty was not.

The UK has no seat at the table. Taken together, this incident, along with this week’s reports of an autonomous AI agent powered by OpenAI models escaping its research sandbox and interacting with the Hugging Face open-source repository, illustrate a profound shift now under way. Our security institutions have been built on the assumption that agency is human. That assumption no longer holds. We are entering a world in which software agents will increasingly make decisions, negotiate with one another and act on behalf of people, companies and nations.

Both incidents report to the same underlying truth—sovereignty is not simply a function of computing power. It is a function of control over access, deployment and consequences, and that is where Britain’s real vulnerability lies.

This Motion is about policy, but we cannot make good policy the way we normally do, through debate, precedent and incremental regulatory convention, when the assumption that those conventions rest on is that intelligence and decision-making are exclusively human. That no longer holds. This is a new magnitude of change. It transforms both the nature of global competition and the meaning of sovereign control.

No single country can dominate every layer of the technology stack, and we should not delude ourselves. UK public investment measured in hundreds of millions can never compete with private capital measured in hundreds of billions. The strategic question for Britain is therefore not whether we can match the raw capability and scale of foreign frontier models but where our genuine advantage lies. Our specific strength is our vibrant British tech ecosystem—in the domain-specific models it is building; the physical systems it is creating, including advanced robotics and drones; and deep vertical applications, on which Britain already holds genuine competitive advantages.

There are risks in that domain as well. Just last week, Microsoft CEO Satya Nadella warned of what he calls a “reverse information paradox”. Organisations that plug directly into foreign frontier models are paying twice: once with their capital for access and again with their institutional knowledge, which they leak prompt by prompt. Palantir’s Alex Karp puts it more bluntly: he calls this losing your alpha—your core intellectual property handed to a competitor that never had to earn it. This reflects a broader shift in where strategic value now resides.

The AI frontier is also commoditising faster than anyone predicted. Intelligence is becoming a utility. Consequently, durable value and, with it, strategic risk have migrated to the layer that interfaces with our critical infrastructure, including our health service, energy grid, transportation and agriculture systems.

The Government’s role should be to set the gold standard by certifying the boundaries around critical infrastructure, ensuring that personal data is protected, data leakages are prevented and British intellectual property is defended. By certifying that protective layer to rigorous standards, we allow our tech ecosystem to flourish safely. We maintain sovereign control over our data, even when the underlying frontier models change, turning digital sovereignty into a competitive asset for British industry.

To achieve this, the Government’s immediate priority must be auditing our dependence on foreign frontier models and establishing strict procurement standards for data sovereignty. That is why I propose a special inquiry into AI power and sovereignty, focused on three non-negotiable pillars: national sovereignty, data sovereignty and cultural sovereignty. Beginning alongside the UK’s G20 presidency in 2027, it would bring together those who build the most advanced AI systems with those who govern them, identifying where Britain’s strategic autonomy is strong, where it is vulnerable and what can be done to protect it. The timing would lead this international conversation with authority at exactly the moment we assume this responsibility. Britain still has time to strengthen its strategic autonomy, but the window is closing. We must not wait until permission is withdrawn to discover what sovereignty was worth.