Why companies should take care when using synthetic data

Financial institutions are under pressure to do two things at once: innovate with artificial intelligence (AI) and protect sensitive commercial and client data. 

Synthetic data is increasingly seen as the solution. It allows firms to train models and test systems without using real customer or market data. 

On the surface, this may appear to reduce risk, making it attractive for credit scoring, fraud detection, regulatory testing and a range of financial modelling applications. 

But our research on synthetic data governance in finance discovered a more complicated reality. Synthetic data does not remove risk, it redistributes it. 

Instead of sitting in the data itself, risk shifts into the systems that generate the data and the assumptions built into them. 

If you overlook this shift, you may end up relying on systems you do not fully understand.

Why synthetic data can increase complexity

The first issue is that synthetic data adds another layer of complexity to already complex systems. 

Financial models are rarely transparent, and introducing synthetic data means there are now two systems at work: one generating the data and another using it to make decisions. This makes outcomes harder to explain and much harder to validate. 

When a model produces an unexpected result, it is no longer clear whether the issue lies in the model itself or in the way the data was generated. 

Using synthetic data as a simple input adds a layer of opacity that can easily go unnoticed, because important choices are made during the data generation process. 

For example, when generating synthetic mortgage data, model developers must decide how much variation or ‘noise’ to introduce. 

Too much can distort relationships between income, repayment history, and default risk, making borrower behaviour less realistic. 

Too little variation may leave the synthetic data too close to the original records, increasing privacy risks. As a result, it becomes harder to trace whether problematic decisions stem from the original data, the process of generating synthetic data, or the model itself. 

Over time, this can weaken model governance and reduce the ability of firms to explain, challenge, and control their own systems. This matters even more when synthetic data is used for validation or testing, as firms may believe they are stress-testing or checking their models when they are actually relying on data produced by another model with its own limitations. 

This can create a false sense of confidence. A model may appear robust simply because it performs well on synthetic data that reflects the same underlying logic.

How synthetic data can create systemic weaknesses

The second issue is less obvious but more serious. Synthetic data can create hidden connections between firms’ data infrastructures. 

If multiple organisations rely on similar synthetic data generation tools, they may begin to see risks and react in similar ways. 

This is how herding behaviour emerges. It does not require firms to coordinate. It only requires them to rely on similar data. 

Over time, this can amplify market movements and increase systemic risk. What looks like independent decision-making may in fact be driven by shared assumptions embedded in similar synthetic data. 

There is also a strategic risk that many managers overlook: loss of control. The process of generating synthetic data involves a series of judgements and design choices that shape outcomes in ways that are not always visible. 

When firms rely on external providers, they begin to lose control of key decisions about how their data is produced. The assumptions that shape their models sit inside systems they do not own and may not fully understand. 

Over time, this creates a form of dependency. It weakens internal expertise and makes it harder to question or challenge model outputs. 

You may still own the model, but you no longer fully control, or fully understand, the data that feeds it. 

Does synthetic data reduce bias?

Another common misconception is that synthetic data reduces bias. In reality, it reflects the data it is trained on.

If the original data contains bias, synthetic data can reproduce it or even amplify it. Because the data appears artificial and ‘cleaner’, it may be trusted more than it should be, creating a false sense of security. 

The issue becomes even more complex when synthetic data is used to correct or rebalance biased datasets. While this can improve representation, it also involves a series of modelling choices about what counts as ‘fair’ and how balance should be achieved. 

These decisions are not neutral. They can reshape patterns in the data in ways that are difficult to detect, especially once the data is used downstream in decision-making systems. 

For managers, this means synthetic data does not solve the problem of bias. It changes how bias enters the system. 

You should therefore test synthetic data just as rigorously as real data, paying close attention to how it has been generated, what assumptions have been made, and how those assumptions may affect outcomes, particularly when decisions impact customers or markets. 

These challenges matter for regulators as much as for firms. Most regulatory frameworks focus on model outputs and decision-making processes. 

But synthetic data sits earlier in the pipeline, at the point where data is generated. This means important risks can be introduced before a model even starts operating. If regulation does not extend upstream, these risks remain invisible.

Why firms needs to understand how data is generated and governed

Policymakers therefore need to rethink how they approach data in financial systems. Synthetic data should not be treated as a technical workaround. 

It is becoming part of the infrastructure of finance. Like credit ratings or benchmarks, it can shape behaviour across firms and create shared dependencies. If a small number of providers dominate this space, it raises concerns about market concentration and systemic vulnerability. 

The practical message for firms is straightforward. Treat synthetic data as a strategic asset. Ensure your organisation understands how it is produced and retains the ability to interrogate it. 

Do not focus only on whether your models perform well. Focus on where your data comes from, how it is generated, and what assumptions it carries. If you cannot explain this clearly, you are taking on more risk than you realise. 

The broader shift is that AI in finance is no longer just about algorithms. It is about data: how it is constructed, shared, and governed. 

Synthetic data is a powerful tool, but it changes where risk is located rather than eliminating it. If you use that data without proper oversight, you risk building  systems that are more opaque, more interconnected, and ultimately more fragile than before. 

As the saying goes: “Garbage in, garbage out.” Without careful governance, the same problem can easily emerge in modelling with synthetic data. 

• This article is based on Spears T., Hansen, K.B., Xu, R., and Millo, Y. (2025) Governing synthetic data in the financial sector. Finance and Society. doi:10.1017/fas.2025.10017

Further reading:

Can generative AI provide better data for financial models?

How to reduce AI bias

Why AI could transform peer-to-peer lending for investors

The flaw in the data that lead corporate bond investors astray

Ruowen Xu is Associate Professor of Accounting. She teaches Foundations of Management Accounting, Powerful Accounting: Critical Perspectives, and Accounting in Practice on our Undergraduate programmes.

Yuval Millo is Professor of Accounting. He teaches Analysing Value Creation on the MSc Accounting and Financial Management and the MSc Accounting and Sustainability.

Discover more about fintech and AI. Receive our Core Insights newsletter via email or LinkedIn.

Choose Warwick Business School as a preferred source of trusted insights in your Google searches.