How Can I Trust My Fake Data?

Synthetic data is increasingly used in healthcare to overcome limitations in access to real world data, enabling innovation while addressing privacy, ethical, and practical constraints. However, its use introduces new risks . In this position paper, DNV presents a lifecycle evaluation approach that supports the realization of synthetic data benefits, emphasizing that trustworthy use relies on structured, transparent, and evidence based documentation, aligned with regulatory requirements, and consolidated in synthetic data cards.

Request a copy

How can we build confidence in synthetic data used in healthcare to support its use for AI‑enabled medical devices, drug development, and guidance of clinical decisions?

AI is transforming healthcare. However it relies on access to large quantities of sensitive data, which are often limited in quantity, quality, and access. Synthetic data provides an opportunity to mitigate these challenges by enabling access to relevant and representative datasets. While it has strong potential to accelerate innovation, synthetic data also introduces new hazards related to representativity, bias, privacy, environmental impact, and compliance.

In this position paper, DNV presents an evidence-based approach to building confidence in synthetic data through structured evaluation of quality. Rather than assuming that synthetic data is inherently safe, of appropriate quality and privacy‑preserving, the paper argues that quality must be documented and decisions justified for each intended purpose and context.

The paper outlines:

• Why synthetic data offers such promise, despite the challenges its use can present.
• How existing healthcare and AI regulations apply to synthetic data in practice.
• A lifecycle‑based quality evaluation framework covering similarity, utility, privacy, fairness, and environmental impact.
• How transparency mechanisms, such as synthetic data cards, support traceability and informed decision‑making.

Download the position paper to learn how evidence‑based assurance can enable responsible use of synthetic data in healthcare, supporting innovation while safeguarding patient safety, trust, and regulatory compliance.

 

Explore the Profiler.SD demonstration

To help illustrate the quality dimensions and lifecycle evaluation approach presented in this position paper, we invite you to explore the Profiler.SD demonstration prototype by clicking this link and entering the password profiler.sd2026.

The prototype provides a clickable walk‑through of how an evidence-based quality assurance‑oriented approach to synthetic data evaluation could be structured in practice, supporting implementation of the key concepts: similarity, utility, privacy, fairness, and environmental impact.

If you would like to provide feedback on the prototype, please contact us at serena.elizabeth.marshall@dnv.com

Disclaimer – Demonstration prototype

The Profiler.SD tool made available in connection with this publication is a demonstration prototype intended solely to illustrate concepts described in the position paper.

The prototype does not allow users to enter, submit, or modify any data, although clicking function is available. All input fields are inactive and presented for demonstration purposes only. No data is collected, processed, stored, or transmitted.

The prototype is not a production system and must not be used for operational, clinical, regulatory, or decision‑making purposes. Access is provided via an externally hosted, password‑protected service.