Artificial Intelligence

Better Risk Management

The new rules establish obligations for both providers and users of AI technologies, depending on the level of identified risk.

In certain areas—such as critical infrastructure management, education, employment, and law enforcement, AI usage must be registered in an EU database. Furthermore, the regulation sets transparency requirements, particularly for generative AI or systems presenting limited risk. “Although many AI systems present minimal risk, adopting this regulation means they must still be assessed,” Eva Gram adds.

Establishing Strong Data Governance

Within this framework, every organization must implement governance associated with AI use. “Artificial intelligence technologies are fueled by company data, so organizations must begin working on data governance now. This governance enables organizations to improve efficiency while managing the inherent risks of data use.”

Data Governance: The Foundation for Compliant and Optimal AI Use

Europe’s agreement to regulate AI represents a significant step forward, providing a clear framework that allows organizations to adopt the technology and benefit from it.

1. A Common Definition
   The European AI law will temper public enthusiasm and provides a clear definition of what constitutes AI. With the widespread access to generative solutions over the past year, AI has sparked discussions, fantasies, and debates—sometimes justified, sometimes not.

This clarifies that AI is not the solution to every business challenge an organization faces. “AI and automation technology are two very different things. Not every chatbot relies on generative AI,” explains Eva Gram. “Generative AI, however, is an advancement based on AI technologies and, as such, requires specific governance.”

2. The Importance of Data Quality Governance
   The regulation emphasizes the importance of data governance and data quality. “If a company wants to use AI, machine learning (ML), or solutions like ChatGPT, it must understand that the quality of results primarily depends on the quality of the data the AI relies on,” explains Arnaud Llagone, Senior Data Scientist at Codit. Poor, incomplete, or outdated data can lead to flawed decisions or operational inefficiencies.

Data governance provides organizations with the capability to manage data effectively. Implementing it requires consideration of regulatory, technological, and operational aspects, alongside developing the necessary expertise to manage and leverage data. Using AI demands raising information security standards, coordinating the DPO (Data Protection Officer), CISO (Chief Information Security Officer), IT, and other roles such as risk management, and documenting data usage efficiently.

3. Developing a Realistic Adoption Path
   AI adoption should be gradual, based on strong governance and aligned with business needs. “AI is first and foremost a tool to support business objectives,” adds Xinxin Li, AI and Data Governance Product Manager. Each organization, depending on its maturity and that of its end-users, must establish a clear strategy, accounting for regulations, IT resources, and organizational structure. Training end-users and implementing a code of conduct for AI use are mandatory under the agreed European regulation.
4. And for Regulated Financial Entities?
   The EU AI Act specifically addresses AI adoption by regulated financial institutions. AI use must comply with other existing laws, particularly GDPR regarding personal data. “With increasing regulatory requirements, the ability of organizations to translate these into IT processes, for example through automated operational workflows, becomes a key challenge,” concludes Eva Gram.

Through its services, Codit acts as a trusted partner to help organizations navigate these challenges and maximize the value of their data. Proximus Luxembourg’s entity dedicated to organizational digital transformation offers consulting, assessment, software development, and training services to support data-driven projects.