As fears mount over AI safety and dependence on US labs, a survey of large UK businesses has found that more than half want greater control over sensitive data, while 29% now view “digital sovereignty” as a board-level priority.
Statistics last week underscored the potential for AI to lift productivity, with expansion in IT services contributing to an unexpected bump in UK GDP growth of 0.4% in July. Yet 58% of enterprises still cite security and privacy concerns as a barrier to AI adoption, according to a survey of 850 multinationals commissioned by Expereo.
Silicon Valley may be to blame. Jacob Coxon, a rank-and-file staffer at Anthropic, last week quit his job and accused the lab of “gambling with our lives” by “racing toward super-intelligent AI”. Three days earlier, OpenAI’s chief scientist warned in a blog of the dangers of an “alien intellect exceeding our own”. In a sign of the internal angst, OpenAI chief executive Sam Altman told staff this week he was “open” to slowing AI development, while the company hired safety expert Paul Christiano to its board.
Beneath these existential concerns, however, are worries that the UK and Europe risk losing out on the economic benefits. At present the continent controls too little of the AI value chain. Given US and Asian dominance in chipmaking and models, could clever use of data be the answer to staying competitive?
“Can we keep our data in Europe if we want to? I think the answer to that is yes,” said Amelia Armour, partner at Amadeus Capital, speaking during a panel discussion at the UK Private Capital summit last Thursday.
But she added that “in terms of the hardware and the underlying semiconductor market, that is much more difficult. Europe is about 11% of [global] semiconductor manufacturing today. It was 10% when we had the EU Chips Act a couple of years ago. The EU Chips Act was €43bn. That’s small potato stuff if you look at what is happening around the world.”
Prior to this week, fears about AI sovereignty reached a peak in early June when the White House placed export bans on two of Anthropic’s latest models, prompting the UK to seek an exemption. Kanishka Narayan, the government’s AI minister, said at the time that the ban highlighted the importance of “sovereign AI capability” and should be treated with “deadly seriousness”.
That episode has prompted efforts to identify a potential “chokehold” that the UK can leverage to retain access to models and compute.
“Is sovereignty self-sufficiency at the expense of relationships? I would say it’s not,” said Katie Ramsay, head of the Global Venture Capital Unit at the government’s Office for Investment. “Having trusted relationships can improve resilience… it’s about ensuring that we have the domestic capabilities for strategically important technologies and the freedom to act if it’s changed.”
In a recent editorial for the FT, Mario Draghi, former prime minister of Italy and former president of the European Central Bank, argued that Europe’s “narrow” zone of sovereignty is to be found in its wealth of public-sector data. But to exploit that advantage will require building large-scale data centres, in spite of public opposition, in order to house data domestically.
Sign up to newsletters from The Observer
For information about how The Observer protects your data, read our Privacy Policy
How would that be achieved without the heft and bulging wallets of American hyperscalers? Draghi’s suggestion is that European companies should pool their demand for data centre capacity, citing an example of ASML, Capgemini and Amadeus, which have together committed to purchases of compute from Mistral, a French AI startup that this week announced a raise of $3bn.
Another option would be to leverage the continent’s deep pool of savings more effectively. “There’s no shortage of capital in Europe – the pension funds could be investing in AI infrastructure development and then selling the compute. It would make a great long-term investment for them,” said Armour.
Photograph by Sameer Al-Doumy/AFP/Getty Images
Click Here For The Original Source
