Last month, a coalition of economists, politicians and AI experts released a report entitled Transformative AI Strategy for Europe. It provides yet more disturbing analysis of Europe’s deep dependency on overseas providers of essential AI. The concept of ‘Transformational AI’ is clearly gaining currency among policy elites as are the dangers it represents.
Co-signatory of the report, ex vice president of the European Commission, Margaret Vestager rightly comments: “We cannot allow Europe to become a playground for foreign powers. Our citizens are already bearing the economic and social consequences of decisions taken elsewhere. AI will accelerate this development, transforming our personal lives and relations, learning and jobs at an unprecedented pace.”
Unrelated to the report, European Central Bank chair Christine Lagarde has also said: “A withdrawal of access [to AI models], or a change in its terms, would … reach every sector at once … That is leverage of a kind no trade partner has ever held over Europe, and it could be used in any negotiation, on tariffs or on digital taxes, for example.”
This is going to be expensive
Yet the solutions proposed by this assembly of experts would exacerbate, not mitigate, the control over European economy and society currently wielded by a handful of ultra-dominant, mainly US-headquartered, corporations.
Two approaches are outlined by the signatories of the report. One is to invest and build European frontier models to compete and participate as equals in the transformation AI economy. This, however, is almost immediately written off as too costly and too high risk to undertake.
The second, strongly recommended route, suggests trading power and land for access to frontier models – Europe will fast-track Big Tech build out on European soil in exchange for “guaranteed” preferential access to their frontier models.
This is little more than capitulation to the dominance of that handful of mega corporations under the control of foreign regimes. It is ironically resonant, in times when we speak of frontier models, of Native Americans trading land for a few trinkets – and we know how well that worked out.
Naïve expectation
It is naïve in the extreme to expect to negotiate lasting privileged access to frontier models in return for power and land in Europe. For one thing – Big Tech already has this access. Europe has a weak hand if trying to extract conditions in return for inward investment to build capacity that it needs.
In addition, current behaviours suggest that corporations and nation states already appreciate and are leveraging the economic and political power that possession of leading frontier models confers. Export restrictions on latest models are already in place, as are mark-ups of 20-50 percent to run models within defined jurisdictional boundaries, like EU clouds.
Recent news unwittingly confirmed that the largest AI developers already guard and retain their best models for themselves. To solve part of the Navier-Stokes equation, a notoriously unsolvable maths riddle, OpenAI used, “a tool only used within the company, as it is “significantly more capable” than the company’s most recent AI model release.”
This clearly illustrates that OpenAI, and surely others, are keeping the best models to use internally. It is therefore inconceivable that Europe, or anyone, will really be able to negotiate access to the true frontier – we would be paying more for a worse model. This is not how you get ahead.
Surrender is not an option
Europe needs sovereign capacity and models to benefit from, rather than become subject to, transformational AI. Europe does not have to accept the Big Tech narrative on what or where the frontier is. Deciding for ourselves how AI can and should be deployed to have the most beneficial impact and best support European values is within our control and is our responsibility.
Regulation gets a bad rap, but one reason we have regulation is to enforce the values we share as Europeans. We care about product safety, fairness, and personal freedom. We do not accept intellectual property theft at massive scale in the name of technological innovation.
We may never catch-up to the US or Chinese frontier Large Language Models, but perhaps we should not blindly accept this as the right race to run. Asking ourselves how can AI positively transform what we already value in Europe must be the first question: what do we really want AI for? Then, perhaps Europe can work on its own large but focused AI models to leapfrog and create AI transformation that matters to us and conforms to our ideals and values.
First, stop digging!
Action is clearly needed, and urgently. But the first step cannot be to negotiate away our assets be they land, power, data or values as part of a surrender to the status quo ante bellum.
The saying goes that when you find yourself in a hole (as Europe most surely is) stop digging. That means, as of now, stop giving new workloads to overseas AI and cloud giants by default.
We already have European alternatives for the majority of cloud services, and for AI, there are both European and open weight models that run with European providers – why are we not using them? Where there are gaps, we can develop bespoke European solutions with European providers. This creates reliable demand for homegrown AI models and compute capacity. It allows sovereign models to train and improve on sovereign data. Crucially, it deprives oversees models from using our data, resources and energy to further cement their control and dominance.
Buy or build European
Nudging public procurement towards European players grows the market and creates reliable demand against which venture capital and corporate funds can invest.
Only by flipping the paradigm – so that the default action is to consider how can we use or create a European solution first – can we fully prepare for the AI transformation and maintain our sovereign control.