In the journey to adopt artificial intelligence, organisations are becoming increasingly aware of the challenges of securing and preserving data confidentiality. For Yannis Nakos, Chief Innovation Officer at NSI, these issues are fundamental and must guide the choices made in terms of the AI environments implemented. In this respect, data control is not an obstacle to the adoption of AI. On the contrary, it should enable lasting value to be derived from it.

Luxembourg companies have no intention of missing the artificial intelligence train. Many of them are now moving from the experimentation phase to the production of use cases. However, this deployment is accompanied by a number of questions, mainly relating to data confidentiality issues.

"There are many risks to consider when using artificial intelligence. The processing of data of a private nature or which is the intellectual property of a company means that a number of precautionary measures need to be taken through the deployment of solutions," comments Yannis Nakos, Chief Innovation Officer at NSI. In the current geopolitical context, with regard to the implications of the US Cloud Act, the simple fact of hosting data with a US cloud provider or processing it via an artificial intelligence SaaS solution offered by a US player implies a risk of loss of control."

The challenge of losing control over your data

Despite a strong European framework for data protection, US regulations allow the authorities on the other side of the Atlantic to get their hands on any information hosted by a US service provider. Processing data using artificial intelligence solutions from cloud platforms is therefore a risk that should not be overlooked.

"The Luxembourg organisations that we support in their adoption of artificial intelligence are aware of these issues," continues Yannis Nakos.

Our role, in this uncertain geopolitical context, is to guide them towards appropriate solutions that offer robust guarantees for protecting the confidentiality of information. It becomes clear very quickly that this is not simply a question of where the information resides.
Yannis Nakos

Yannis Nakos, Chief Innovation Officer, NSI

When data sensitivity drives choices

The idea is not to demonise "made in the USA" solutions. In fact, for NSI's Chief Innovation Officer, that would be counterproductive. Good luck finding a non-US operating system," he says. Many solutions from the United States are efficient, appropriate and meet the needs of European players."

The challenge lies more in the way the data is stored and processed. "Before deploying artificial intelligence solutions, each player should ask themselves where their different data is stored, given the sensitivity of each of them. Are they hosted in their own data centre, in a sovereign cloud or with an American cloud provider?", asks the specialist.

Beyond that, you need to be able to identify the data you want to subject to artificial intelligence in the context of the processing and use cases envisaged.

AI models tailored to business needs

"Depending on requirements, it is a question of determining how AI solutions should be deployed. Artificial intelligence models, whether European, American or Asian, can be implemented from a sovereign infrastructure, like the one NSI has implemented in Luxembourg. In this way, the organisation is assured that no American authority is able to get its hands on the company's data," explains Yannis Nakos.

The choice of AI model, in this context, will depend on the customer's need. Today, organisations can access a wide range of open source models that can be easily deployed from our sovereign AI cloud," continues the Chief Innovation Officer. Numerous pre-developed agent models and solutions offer measurable gains in productivity, service quality and analytical capacity, particularly in cross-functional areas such as HR, marketing and IT support. The technology is operated within a neutral framework. The data remains in the customer's environment. In this way, they have guarantees that their data is not at risk of being exposed to third parties, due to foreign regulatory provisions."

Between security and costs: Strategic trade-offs to be made

Deploying AI models and services in a dedicated environment in order to guarantee the security of processing does, however, involve higher costs than those of widely shared solutions. "However, it is important to assess the costs as well as the risks. Depending on the uses and data to be processed, organisations will opt for solutions based on the public cloud or others operated from their own environment," continues Yannis Nakos.

We are increasingly seeing the emergence of solutions, such as the Sovereign Cloud software platform published this month by IBM, for example, and adopted by the NSI Group, enabling sovereign AI environments to be deployed rapidly and completely securely, in full compliance with the European IA ACT. They allow organisations to test, iterate and industrialise their use cases without compromising the security of their data or slowing down their transformation.

Governance as the foundation for sustainable adoption

"Our role, to help organisations adopt AI is to enable them to make informed choices, based in particular on robust data governance that complies with current European regulations", concludes NSI's Chief Innovation Officer. Clear data governance not only reduces regulatory risks, but also speeds up the production of AI use cases, with better alignment of costs, appropriate security as well as IT, business and legal teams.