CURRENT EDITION

Synthetic data: legal, ethical and technical implications

May 28, 2025 - 11:00 - 12:30

Room 2 (Σ123)

Rational

Synthetic data can effectively supplement or replace real-world data in training and evaluating AI models, especially when real data is scarce, sensitive, costly, or difficult to label. When generated with deep domain knowledge, synthetic data can be produced at scale, offering benefits such as class balancing, rare event simulation, and improved privacy.

However, challenges arise when synthetic data fails to fully represent real-world complexity. These include risks of over-smoothing, missing temporal or dynamic patterns, and omitting rare but critical outliers—potentially leading to biased or inaccurate models. Validation against original data is often required, particularly for complex, high-dimensional datasets.

The integration of large language models (LLMs) in synthetic data generation has advanced capabilities in data-scarce settings but also introduces concerns about bias amplification, representational fairness, and data fidelity. Ensuring diversity, accuracy, and equity remains essential for trustworthy AI outcomes.

Objectives

This interactive workshop aims to explore and discuss the legal, ethical and technical implications raised by the creation and use of synthetic data, including a keynote presentation of BDVA Healthcare TF.

The expert speakers in the workshop will investigate and discuss the legal, ethical, and technical implications, including opportunities and challenges, of generating and using synthetic data, engaging with the audience with different interaction strategies and tools. Special attention will be given in seeking for a careful balance between, on the one hand, the synthetic data generation, leveraging the benefits of domain knowledge and real-world dataset examples, and on the other hand, respecting the confidentiality and privacy of individual and upholding European values. The speakers and the participants will also discuss about the responsible use of synthetic data in AI, and the need of a robust and legal compliant policy framework along with ethical guidelines which should accompany the development and use of these technological solutions.

This workshop will invite and engage many participants as this is part of the activities of HELEN (Horizon Europe Legal and Ethics Network, https://www.helen-hub.eu/). The network currently encompasses already 28 European projects, with additional projects and initiatives expressing interest in participating. HELEN is designed to address the legal and ethical challenges arising from recent technological progress across diverse sectors, including advancements in Data, Artificial Intelligence, and Robotics. Projects AI-DAPT, AI4Gov, ATLANTIS, CyclOps and SYNTHEMA will contribute to the workshop.

Agenda

The agenda is structured in two main parts.

Introduction of the key workshop topics.
Moderator: Dr. Germana Gianquinto
- Technical Overview on Synthetic Data: Generation methods, Assumptions and Challenges by Sotiris Koussouris, Suite5 Data Intelligence Solutions Ltd, Cyprus – AI-DAPT Project
- The key legal and ethical implications of using Synthetic Data by Marina Cugurra, R&I Lawyer and Ethics Expert, S&D Consulting Europe, Italy – AI-DAPT Project
- Presentation of the White Paper of TF Healthcare BDVA by Sofia Tsekeridou, Netcompany – Intrasoft – SYNTHEMA Project & BDVA TF Healthcare (Keynote Speech)
- Bias identification and mitigation strategies in synthetic data: insights from the eGovernment domain by Konstantinos Mavrogiorgos, Department of Digital Systems, University of Piraeus, AI4GOV Project
- Harnessing synthetic data and AI to improve financial services by Konstantina Tripodi, JRC Capital Management, Germany – ATLANTIS Project
Interactive panel with:
- Argyro Amidi and Alexandros Kalafatelis, representing Four Dot Infinity and the CyclOps Project;
- Saurav Baidya, representing Philips and Co-lead of the BDVA TF Healthcare.