The phrases ‘Data Sovereignty’, ‘Digital Sovereignty’, and ‘AI Sovereignty’ are coming up more and more, so I thought it would be useful to try to unpack them, and understand why they matter. The terms sound similar, but I’d argue that they are in fact, quite different, with very different concerns.
Data Sovereignty is concerned with control over data, including where it is stored and processed. If you look for definitions you find some variance, with some looking at it from a national perspective, and the ability of a nation to govern the data within it’s borders, and some looking at more from an institutional perspective and it being about the ability to control access and use of your data. For example, it might include the ability to store and process data (data residency) in a geographical area so you are happy from a legal and compliance perspective, and controlling who has access to the data.
Digital Sovereignty discussions are usually much more about individual nations or groups of nations (such as the EU) to have control over their digital destiny, typically through access control of digital infrastructure. As a discussion, it’s relatively recent – for example the EU outlined their need for digital sovereignty in 2020.
AI Sovereignty discussions are usually a subset of Digital Sovereignty and have become a pressing issue as generative AI has become dominated by a small number of companies, in the western world mostly concentrated in the United States. They are concerned with a nation’s ability to have control over its use of AI through infrastructure and skills. Vendors discuss this, and you may view it as a means to sell technology (see Nvidia and Oracle), but it’s also increasingly part of national strategies.
Why does it matter?
You might ask why does it matter? And hasn’t technology always been provided by a small number of players?
International supply chains are of course complex, and we’ve all seen recently what a huge impact it had when, for example countries relied on another country for the supply of gas, and relationships with that country changed.
You might say that this isn’t going to be an issue for AI, and that surely friendly nations will just continue to supply services as it’s in their economic interests. But we are already seeing that this isn’t the case, and besides, as we can see, situations can change very quickly. For example, the approach of the USA might change dramatically with a change of leadership.
Let’s pick up some real examples. Almost all of these are related to regulatory differences between the EU and the USA, and they can happen in both directions.
- In April 2023 the Italian Government briefly banned ChatGPT due to concerns about compliance with GDPR. All access was lost for a while.
- In June 2024 Apple announced that their AI tools wouldn’t be made available in the EU due to concerns about the EU Digital Markets Act (DMA)
- Google Gemini image generation is not available in the EU, Switzerland and UK. They haven’t specified why, but it’s almost certainly related to regulation
- AI at Meta is not available in Europe, but is in many other parts of the world, due to regulatory concerns. This includes their cutting edge multimodal models, which likely will be available in the UK, but not the EU.
The lesson here is clear – a country or bloc can be fully in control of its legislation, but that doesn’t amount to AI sovereignty. It doesn’t mean that overseas companies will chose to provide services to that country, so doesn’t give the country control over its AI future if it is reliant on services from international partners.
If we look at international landscape, the mainstream AI players we are familiar with like OpenAI, Google and Meta are American. China has a rapidly growing generative AI sector, The EU has a significant generative AI company in French company Mistral. It might not be a household name, but’s it’s generative AI models are highly capable. The UK has no such equivalent.
Should we have a UK based Gen AI solution? I certainly don’t think it would be a bad thing, and it’s been proposed by Labour for the Long Term. The new UK governments commitment to green energy is probably one of the key building blocks. As we know, one thing AI needs is energy, and it’s in everyones interests that that energy is green and low cost. I’d argue that we’ve already got the skills base.
You might say that the answer to this is use open-source AI, and in an ideal world that might be true, but I’d argue that today most of the mainstream open-source AI solutions, certainly from the big players, don’t provide options for medium to long term AI sovereignty – I’ve written a separate post about that, but it’s largely because they control the training, training data, and training infrastructure, and also hold the skilled workers to develop the model.
So, what now?
You might well think that this doesn’t really have much to do with Universities and Colleges using generative AI in education, and in some sense that’s true. I’m certainly not arguing that we stop using these tools until we have AI Sovereignty. It’s something we should be aware of, and include in our broader AI literacy content. You may also want to consider it in your risk processes, depending on the detail that your institutions goes into, and consider the impact of vendor lock-in, and what the mitigation would be if AI features from our preferred vendor weren’t made available to us, but, say, to our global competitors. We probably shouldn’t lose sight of the fact that although we aren’t part of the EU, issues around EU regulation still impact us, as from a commercial perceptive, vendors may well just choose to consider Europe as a whole
And finally, we should also contribute to discussion with government and other policy makers.
Find out more by visiting our Artificial Intelligence page to view publications and resources, join us for events and discover what AI has to offer through our range of interactive online demos.
For regular updates from the team sign up to our mailing list.
Get in touch with the team directly at AI@jisc.ac.uk