The world is currently in a state of unprecedented data availability. Within the financial sector, and indeed every other sector, organisations have found themselves with an abundance of data and information that they can use to better navigate the changing face of commerce. In many ways, this is incredibly helpful. Data is often a major catalyst for technical innovation and digital transformation. Advances in data analytics, and an increase in abundance, has allowed organisations to automate various processes, increase oversight of the industry with which they are involved, and reduce potential risk vectors.
Innovation in this form is particularly prevalent in the financial services sector (“FSS”). For example, consumer interaction with financial products, and the data these interactions output, have allowed financial service providers (“FSP”) to develop digital footprints of their customers. This allows them to tailor and personalise the service they provide to better meet the needs of those they serve. The drawback to this use of technology is the vast quantities of recent and relevant data it requires. As with many personal data points, financial data is deemed sensitive and subject to a number of privacy laws (for example, the GDPR and the Data Protection Act 2018). While there are ways of accessing and sharing this information in compliance with regulation, the process is often time consuming and full of hurdles.
In a bid to overcome this type of issue, the Financial Conduct Authority (“FCA”) have, through initiatives such as their TechSprints, sought to find new ways of enabling innovation. One potential method they are considering is the use of synthetic data to replace many instances where private data is currently used. In order to further understand this option, the FCA has issued a call for input from FSS stakeholders on their thoughts on the matter, and whether synthetic data could have a place in the FSS moving forward.
This article briefly details what exactly is synthetic data, the FCA call for input and what they seek to discover, and how organisations can have their say moving forward.
What exactly is synthetic data?
So what exactly is synthetic data? The Office of National Statistics defines this as:
“microdata records created to improve data utility while preventing disclosure of confidential respondent information”.
In essence, it is data that has been artificially manufactured, rather than becoming available from a naturally occurring event, such as a customer entering their personal preferences of banking products. This is done by detecting patterns of data behaviour or trends in real data and then using AI to produce artificial datasets that align with the actual real information.
The primary benefit, as noted by the FCA, is that FSS organisations can leverage the overall trends of their customers (or those materially similar to those in behaviour) without having to deal with the added hurdles of privacy regulation. The compromise with this data is that it is only as good as the original data used to model its creation. Failure to provide a reflective training input or utilising insufficient amounts of source data may lead to a biased data set that does not reflect the market (thereby rendering the synthetic data of little value, or even worse making it misleading). Synthetic data does not therefore act as a ‘deus ex machina’ with respect to privacy and insufficient data, and suitable care should therefore be taken in its use.
Synthetic data offers FSPs a number of opportunities to develop and share their products and expertise without the added restraints that come with the use and distribution of actual data. It is therefore of little surprise that the FCA deems it one of the most promising tools for using and sharing data in situations, such as financial services, where privacy is of great importance.
The FCA call for input:
The call for input broadly covers three main subjects:
- data access and innovation;
- synthetic data and the financial services; and
- potential roles for financial regulators.
Data access and innovation:
The FCA recognise that for many firms, financial data is key to them developing new and innovative products and solutions. This is particularly so in the instances when AI and similar ‘data thirsty’ technologies are used in order to do so. Therefore, in order to continue the development of products, such as in relation to financial crime and fraud prevention software or asset management optimisers, ensuring sufficient access to appropriate data is imperative. As such, the FCA have requested information on methods of making available otherwise sensitive data and whether organisations are currently facing significant issues accessing the data they need in order to innovate.
Synthetic data and the financial services:
It is clear that the FCA view synthetic data as a strong enabler for innovation where privacy restrictions and lack of real-world data exist. This is not without reason. Synthetic data is already in use in a number of other sectors and situations. For example, financial models on emerging markets are being used to plug the gaps in information that traders have when making investment decisions. More tangibly, synthetic data has also been used in the training of driverless cars, prior to their being placed on public and private roads.
As such, the FCA have highlighted three “overarching benefits” to the use of synthetic data:
- data privacy: as the data being used should not be capable of being attributed to an actual person, there are no concerns that the privacy of customers or contributors will be infringed;
- plugging the gap: in similar fashion to the examples above, the FCA envisages that synthetic data can be used to plug the gap where no or little data exists, such as training new counter-fraud programmes; and
- cost efficiency: large volumes of real-world data are often costly to obtain due to the difficulty it takes to generate. Synthetic data, due to its ease of creation, is much more cost effective with which to train and test programmes.
The FCA have therefore dedicated much of the call for input to this section. In particular, respondents are asked to consider methods of how this data can be generated, how it can be fully leveraged in the financial services, and how this can all be done while ensuring privacy and mitigating risk.
Roles of financial regulators:
The final aspect of interest to the FCA in the call for input is how regulators should be involved in a way that meets all of their operational objectives (market competition, market integrity, and the protection of consumers). They are therefore considering market attitudes towards synthetic data and where they should sit within its use.
The FCA propose three potential roles regulators could play moving forward (none of which need to be mutually exclusive):
- Data Generator: In this role the regulators would collaborate with industry experts and academia in order to generate accurate synthetic data in-house. The regulator could obtain real data from various sources, in order to limit bias, and then create their own set that can be distributed across all sectors.
- Central Host: In this role the regulators provide an independent hosting platform for the storage of synthetic data. From here organisations can store and share data to be used for the purposes of development and testing of products.
- Coordinator: In this role the regulators would act as a coordinating middle body to facilitate the sharing of data in a safe and efficient manner.
All roles would require cooperation with organisations of all sectors and market opinion on the openness to do so, alongside any practical concerns, is particularly requested.
How can I get involved?
The FCA is keen to hear input from all stakeholders of the FSS and will accept contributions until 22 June 2022.
Respondents are requested to reply to the points detailed in Annex 1 of the call for input and email their conclusions to: firstname.lastname@example.org
The results of the call for input will be reviewed and published in due course.
DLA Piper continues to monitor updates and developments of AI and its impacts on the financial services sector. For further information or if you have any questions please contact the authors or your usual DLA Piper contact.