The Ethical Use of AI in Today’s FX Marketplace

This is the second in our occasional series, The Voice of Experience, which taps into the experience of senior, and independent, FX market veterans, and invites them to discuss an issue in the FX industry that maybe others prefer to avoid.

This edition, Steve Flanagan asks, should we be concerned about the ethics around the use of AI in today’s FX market?

The Ethical Use of AI in Today’s FX Marketplace

AI has been used in the FX world for at least the last five years and its use continues to grow with the advancement of machine learning (ML). It drives market making, pricing, risk management as well as direct client tiering, and will continue to work its way through more and more front and back-office functions, streamlining jobs and driving efficiencies.

The key to just how well AI interacts within a franchise is the amount of data available to the business that will enable it to learn and build its business, and herein lies the risk to the FX industry. Are potential future regulators likely to look back and question how that data was used and how AI and ML were applied to the marketplace?

First, though, a brief look back. We all know the ethics lapses that our industry suffered in a decade or more ago, specifically involving the chat rooms. But it was the breakthrough in data storage that shed light upon the growing use of those communications as well as automated pricing. It meant an electronic footprint was saved and stored forever. This meant the regulators could not only ascertain motive, but they could also figure out the subsequent action that those they were investigating took. For the first (but not last) time, the data of saved chats were used to validate charges of unethical behaviour.

It is also probably worth wondering what would happen if one were able to go to the reel-to-reel tapes of the conversations in the 1980s or the miles and miles long Reuters Dealing chats? How would that magnifying glass of ethics today look to the markets in pre-stored data days? Not well! That is because the foreign exchange markets grew out of a history of the exchange of information – a true trading relationship. Don’t frown at that comment, one only needs to remember the days of central bank intervention after the Plaza Accord in 1985 and being asked by the central banker doing the intervening, “who’s buying the dollar?”

The central banks were selling and wanted to know who was buying? This, and of course who is selling, were questions asked each and every day to a liquidity provider. There was little price transparency in those days, no order book information and very little in the way of price history, therefore customers, the corporates, hedge funds, asset managers and central banks, all wanted that information to help make a better trading decision. So, the information was exchanged and it was normal, but how does it look now?

It is nothing new knowing a customer’s expected tendency, but in an age of transparency with more data than ever at hand – and easier to access – is going to look ethical?

The maturation of the industry after 2000, especially as automation levels rose and electronic trading became prevalent, meant that the information demanded by those clients in the 1980s and 90s was more readily accessible. As noted, though, this also shed light on the few who used that exchange of information in a negative fashion.

The positive today is that we all are subject to the guidelines, rules and regulations that arose from the industry’s mistakes. We have advanced our self-regulation using bodies such as the FXPA and ACI FMA as tools to engage regulators and policy makers and thus advancing a transparent and competitive global currency market, which benefits us all.

We should remember though, that it was a painful process to get to this stage and as such we need to as the question ‘Are we prone to make mistakes again?’ and will tomorrow’s regulators view our actions through a very different lens?

The key issue will be how the market data, which lies in the hands of even fewer LPs, be used and what will tomorrow’s regulators think about it? For while the data has become more transparent, and all can view them through pre- and post-trade analytics from both liquidity providers and, more recently, third-party venders, where does AI and ML take our use of it?

Is there a potential pitfall in how, as technology advances even further, those few LPs, with the most market data, begin to learn and shape their liquidity offering? Why shouldn’t they, after all? Post-trade that data resides with the LP and risk management is an LP decision.

Today the FX industry is in a good place, where market transparency benefits all, but it needs to advance carefully

It is when the shaping of price takes on a pre-trade view that years down the road, regulators may look back, through a different lens, and say “you can’t do that.” It is nothing new that studying a client’s trading behaviour and knowing their expected tendency puts an LP at a distinct advantage, but in an age of transparency with more data than ever at hand – and easier to access – is going to look ethical? It could be, as we’ve all done, the equivalent of opening a search engine and seeing multiple ads for an item we have recently been shopping for elsewhere on the Internet, luring us into a purchase. Or it could be seen in a more sinister light (as the search engine’s behaviour is often viewed!).

Today the FX industry is in a good place, where market transparency benefits all, but it needs to advance carefully. Ethics in the near future could fall back to the developer who programmed the response to the data and if the primary directive input to the AI is profit, then we can all just imagine where a machine that learns from market data, which is available to a few, could potentially go.

Let’s ask the questions today and not wait for tomorrow’s new ethical judgement look-back.

Steve Flanagan traded FX for more than 40 years, the majority of it in banking groups now under the JP Morgan umbrella. He retired from JP Morgan in 2022 from the role of global FX e-commerce risk manager.