AI in Markets: Embedded, but not Autonomous

The latest Spotlight Review from the Financial Markets Standards Board (FMSB) offers an insight into AI and how far it has permeated the market landscape, concluding that while it has become more embedded in processes, it is yet to operate autonomously in market-facing environments.

The paper, AI in Trading; A Practitioner’ View of the Current Landscape, rather than try to define AI – which it observes is “inherently challenging” thanks to the pace of development, something that can render definitions outdated very quickly – offers observations on a spectrum of simple to advanced techniques. It does this through several use cases, that FMSB says, “raises considerations distinct from those associated with well-established quantitative methods”, most notably those already embedded in existing control or governance frameworks.

Equally, while generative AI techniques “may warrant future consideration” as the technology evolves, specifically into direct trading applications, this too, is outside the bounds of the latest Review.

The paper offers case studies on deployment in wholesale markets and model risk management, as well as look at the potential systemic risk considerations from its use. It also looks at where AI operates in e-trading systems, its use by market making agents, and in CRM (customer relationship management) systems with specific focus on pricing to those clients. Each case study offers a synopsis and a scenario, alongside the potential risks and impacts to the business.

While emphasising that despite growing sophistication, market‑facing AI does not currently operate autonomously, instead it is embedded within existing trading infrastructure and remains subject to direct and indirect human supervision, supported by established algorithmic trading and model risk controls, FMSB warns that institutions need to remain focused on the potential issues as it advances further. 

“Existing control frameworks already manage many AI-related risks well, but they may need to be kept up to date as the scale and complexity of the application of AI in markets further develops,” the board states. “Clear human accountability remains essential for decisions made by AI in trading, just as with traditional systems. While autonomous AI in trading could emerge in the future and pose systemic risks, such fully autonomous trading systems are not yet present in financial markets.”

It further notes, however, that “the scale, speed, or novelty of certain AI applications may place pressure on existing guardrails and give rise to additional considerations to which market participants should remain alert”.

David Shelton is global head of the FICC electronic trading and FX quantitative strategies group at Bank of America and chair of the FMSB AI Working Group that produced the Spotlight Review; he says, “We hope it will give firms a useful practitioner perspective to help harness and deploy AI in market-facing applications in a controlled way, using existing risk management frameworks while recognising their limits. As a Working Group we intend to continue our work in this area, with a view to establishing specific good practice guidance around model risk management and trading controls for AI models.” 

Myles McGuinness, FMSB CEO, adds, “AI has the potential to transform the functioning of wholesale markets, but its benefits can only be realised if firms pair innovation with robust governance and accountability. This Spotlight Review provides a timely insight into how AI is currently being used in wholesale markets and the opportunities and risks at a time when market participants are actively exploring how to deploy it safely and effectively.

“Ensuring human oversight, strong controls and clear responsibility and accountability will be essential to sustaining fair and effective markets as the application of these technologies evolves,” he concludes. 

The full paper can be accessed here