In this ELOQUENCE Web Café session, the focus was on a topic that often remains in the background of AI development, but plays a crucial role in shaping responsible and trustworthy technologies: data governance.
The session featured Anastasia Kiseleva and Aayushi Gupta from Privanova, who brought legal, ethical, regulatory, and project management perspectives to the discussion. Together, they explored why data governance is not simply an administrative requirement, but a central pillar of responsible AI, especially in EU research and innovation projects.
Anastasia opened the conversation by emphasising a simple but essential point: without data, there is no AI and without good data, there is no good AI. Trust in AI systems depends heavily on the data they are built on. If data is incomplete, biased, irrelevant, or poorly managed, AI systems can produce harmful, unfair, or discriminatory results. This makes data governance essential for protecting fundamental rights, ensuring transparency, and building public trust.
In the European context, data governance is also a regulatory requirement. Anastasia referred to the EU AI Act, particularly Article 10, which sets requirements for training, validation, and testing datasets used in high-risk AI systems. These requirements focus on data quality, relevance, representativeness, bias prevention, and proper documentation.
To make the concept more concrete, Anastasia explained data governance through several practical principles: knowing the context in which the AI system will be used, knowing the data and its sources, understanding the assumptions behind it, managing risks, identifying gaps, documenting processes, and clearly defining roles and responsibilities. In other words, good data governance means being able to explain not only what data is used, but why it is used, how it is processed, and who is responsible for each step.
Aayushi then brought in the project management perspective, explaining how data governance affects the day-to-day coordination of large EU consortia such as ELOQUENCE. In practice, it shapes project discussions, partner alignment, documentation, risk management, reporting, and validation. Since different partners may have different internal policies or interpretations of safe data sharing, clear governance helps avoid confusion and ensures that collaboration remains both efficient and compliant.
The session also highlighted that data governance should not be seen as something that slows innovation down. On the contrary, when introduced early, it protects innovation by making project outcomes more credible, transparent, and aligned with European standards for responsible AI.
Looking ahead, both speakers pointed to the need for data governance to become more embedded, continuous, and practical. Rather than being treated as a final compliance check, it should be part of the project from day one. It should also evolve from a company-level practice into an ecosystem-level approach, especially in collaborative projects where data moves between multiple partners.
The key message was clear: data governance is not just about compliance. It is about creating the conditions for AI systems that are trustworthy, accountable, and genuinely aligned with human values.
Watch the full session here.
