The UK’s Council for Science and Technology published advice to AI minister Kanishka Narayan on October 19, setting out four recommendations aimed at what it calls AI sovereignty. The letter itself is dated July 30.
It is signed by co-chairs Dame Angela McLean and Lord Browne of Madingley, and lands at a moment when control over models, compute and data has become a live political question across Europe.
The gap between advice and capacity
The council’s framing treats sovereignty as more than hosting foreign models on domestic servers. Building and directing capability implies domestic research, domestic compute and the ability to set terms for how AI is used inside the country.
Britain’s position is awkward. It hosts a strong research base and several frontier labs, but relies heavily on US hyperscalers for compute and on imported models for deployment. A government that can regulate AI it cannot produce has leverage on paper and much less in practice.
The four-month gap between the letter’s date and its publication is worth noting. Advice written in July reached the public in October, after a summer in which compute access became a bargaining chip in trade discussions and export-control debates.
Earlier in October, Nvidia committed $2B to Europe and handed UK labs free compute, an arrangement that improves access while deepening dependence on a single supplier. That trade-off sits at the centre of the sovereignty argument.
The council’s recommendations are advisory. Whether they survive contact with Treasury spending rounds is a separate question from whether ministers accept the logic.
