Three new signals sharpen the synthetic data conversation: a major ERC-funded project will interrogate societal impacts, the World Economic Forum publishes a governance-oriented playbook, and new medical evidence quantifies privacy-versus-utility trade-offs in high-dimensional datasets.
New project to investigate societal consequences of using synthetic data to train algorithms
The University of York announced the launch of SYNDATA, a European Research Council-funded research project led by Dr. Benjamin Jacobsen. The project will examine the practical, ethical, and political consequences of using synthetic data to train algorithms across sectors including healthcare and finance.
SYNDATA plans to use archival research, fieldwork, and case studies to analyze how synthetic data affects society and power structures—especially as generative AI makes it harder to distinguish “real” from “synthetic” data in the AI supply chain.
- Governance is moving upstream. If SYNDATA surfaces repeatable patterns of harm (or benefit), expect more scrutiny on how synthetic datasets are produced, validated, and documented—not just how models behave at deployment.
- Procurement and audits may broaden. Data teams could be asked to evidence the social and political implications of synthetic data use (e.g., who is excluded, which assumptions are encoded), beyond standard privacy and accuracy checks.
- Standards pressure increases. Findings could feed into regulator expectations and cross-border norms on data ethics and algorithmic fairness as synthetic data becomes a default workaround for access constraints.
Synthetic Data: The New Data Frontier (WEF briefing paper)
The World Economic Forum published a briefing paper positioning synthetic data as a scalable response to data scarcity and privacy constraints, particularly in sensitive domains like healthcare and finance. The paper highlights use cases including testing, personalized AI, and red-teaming, and frames synthetic data as a practical tool to unlock AI development when real data is limited or too risky to share.
At the same time, the WEF emphasizes the need for governance to manage accuracy, equity, and privacy risks—implicitly acknowledging that synthetic data is not automatically “safe” or “representative” just because it is generated.
- Expect “synthetic-by-default” proposals. The WEF framing will be used internally to justify synthetic data programs; data leaders should be ready with criteria for when synthetic is appropriate versus when it degrades validity.
- Control objectives are becoming standard. Accuracy, equity, and privacy are presented as governance pillars—useful as a checklist for internal policies, vendor evaluations, and model risk management.
- Collaboration becomes a compliance tactic. The paper’s call for multi-stakeholder coordination signals that regulators and industry groups may converge on shared assurance practices (documentation, testing, and monitoring) rather than bespoke one-offs.
Impact of synthetic data generation for high-dimensional cross-sectional medical data: privacy versus utility considerations
A new study in the Journal of the American Medical Informatics Association (JAMIA) evaluates three synthetic data generation (SDG) strategies for high-dimensional, cross-sectional medical datasets. The authors compare privacy risk—specifically membership disclosure—against downstream utility, contrasting approaches that generate synthetic data from the full dataset versus subset-based strategies.
The paper offers evidence-based recommendations aimed at data sharing platforms that need to balance privacy protection with analytical usefulness for AI-driven healthcare research.
- Privacy claims need measurement. The study’s focus on membership disclosure reinforces that “synthetic” does not equal “non-identifiable”; teams should quantify risk rather than rely on labels.
- Utility can be engineered—at a cost. Full-dataset versus subset generation choices change the privacy/utility balance; platform operators should treat SDG configuration as a governed decision, not a default setting.
- Regulated sharing gets more concrete. Evidence on trade-offs supports defensible design choices under GDPR and HIPAA-aligned controls, especially when synthetic data is used to enable broader access for research and development.