TL;DR

Perplexity has signed a partnership with Valarian, a London start-up, to bring its AI into British government and defence systems, City AM reported on 20 October. Valarian’s software will let organisations with sensitive data run Perplexity inside infrastructure they control, keeping a say over where their data goes and how the model runs. AI minister Kanishka Narayan hailed it as a sign of what British firms can offer.

How the arrangement works

Valarian sells a control layer: its software decides which AI models may reach an organisation’s systems, which data they can draw on, and where each workload runs. Under the deal, a customer could send general research queries to externally hosted AI while keeping more sensitive work on isolated infrastructure it governs itself.

Perplexity, which says it has more than 45 million users worldwide, expects the deal to open up sectors where data security has held adoption back. Nate Kupp, who leads enterprise and infrastructure engineering at Perplexity as a vice-president, put it this way: “What’s held them back is where it runs and who controls it.”

Narayan said: “When one of the world’s leading AI companies needs to deliver sovereignty to enterprise and government, it is choosing a British company to do it.” He added that critical industries and public services should be able to use AI “without handing over control of how and where it operates”.

Max Buchan, Valarian’s founder and chief executive, told City AM that without control of their data, organisations risk relying on infrastructure they do not govern, or on AI systems “that could leak their intellectual property”.

Context

Valarian raised a £37 million ($50 million) Series A in July, led by US investor New Enterprise Associates, which took its total funding to about £52 million ($70 million); we covered that raise at the time. City AM says its technology is now in use on four continents.

The deal lands in the same week as a warning from the government’s top science advisers that Britain could end up a “high-value AI customer” rather than building its own capability. The government is also pressing ahead with its £500 million Sovereign AI Fund.

Looking forward

In our view, the deal shows one version of sovereignty that UK buyers can adopt now: a US model, governed by British software, running where the customer decides. That differs from building domestic models from scratch, and the two are not interchangeable. For public bodies and regulated firms, the practical question is whether a control layer of this kind satisfies their own data-handling rules, and that will need testing in procurement rather than taking on trust.