Few things in life are met with both optimism and scepticism. But then again, even fewer things in life can be compared to artificial intelligence. On the one hand, AI is viewed as a revolutionary force capable of transforming industries, while its potential is still largely untapped and unknown. The truth is the only word that accurately describes AI’s potential impact is: ‘uncertainty’. We just don’t know which way it is going to go, and uncertainty breeds fear.
This uncertainty is clear in trading environments. The increasingly prevalent use of AI has led to industry experts raising concerns about its role in financial services, primarily because of the unknown factors it introduces and how the decisions it makes can be attributed. Regulators now find themselves with the challenge of trying to distinguish between legitimate trading and non-compliant activity that is underpinned by AI algorithms and decision making. This is a particular worry in cases where potential collaboration between AI-powered trading systems could lead to widespread, high-frequency manipulation amongst multiple actors.
AI has also amplified existing worries about bias and misinformation. The technology has increased the potential for individuals with minimal trading or regulatory knowledge to enter the market, potentially increasing the complexity of managing and regulating these activities as part of an ever more interconnected market.
However, is the concern surrounding the use of AI in trading truly justified? And is the industry facing a radical shift that will change its foundations forever?
A reality-check on AI risk
On closer examination, the answer appears to be no. Reflecting on the introduction of algorithmic trading around 15 years ago provides valuable perspective and was, in many ways, a more seismic change. Algorithmic High Frequency Trading (HFT) allows firms to execute trading strategies in milliseconds, exponentially quicker than traditional, slow, human traders could enter and execute orders.
AI is not improving on these speeds (and has not yet opened up new forms of manipulation created by relative speed or market impact changes) as it operates in a fundamentally similar fashion – the risk it does create is that of collusion between different AI systems for the prospect of mutual gain.
How do you mitigate the risk of AI actors behaving in ways that are not directly scripted? The answer is to understand how manipulation occurs and build best-in-class systems to identify and report upon instances of potential abuse – to almost be agnostic to the trader itself.
The business standpoint
It’s a firm’s role to have oversight of their risk. When it comes to traits of variance in AI, firms need to own their risk and understand why controls are in place and how they operate. Is AI going to manipulate the market because profit making is defined as the key outcome from its trading activity, for example? Will an AI trader override controls put in place in order to achieve a primary (profit making) objective? How do you manage a trading system that is deliberately designed to execute with a degree of initiative that goes beyond current, algo-driven and rules-based, systems?
Unmanaged AI is a risk. But controlled, managed, and mitigated AI is no different than having a series of parameters that have been tested and tuned by a human. Ultimately, you need to understand how you’re controlling your systems and your processes. It’s also important to note that managed AI should be considered as a new toolset that can enhance how you deal with problems and imagine solutions for them.
We’re very unlikely to get to the stage where generative AI runs the marketplace; that’s not a likely risk at the moment. Ultimately, the person who is in charge of releasing the AI algo into the wild has to realise that they need to engage in some form of manual control that sits above it. And that’s no different from algorithms and HFT with RTS-6.
The regulatory standpoint
New forms of technology like AI need regulation. And there is often much written about how regulators need to preempt criminal activity and emerging types of market abuse. However, regulators cannot preemptively ban or regulate forms of abuse that have not been invented yet. The way to avoid regulatory gaps is to create a regulatory environment that covers a combination of both specific behaviours and principle-based requirements.
With robust controls and regulatory frameworks in place, it is far easier to tackle new threats effectively. This requires a collaborative ecosystem where regions, regulators and firms are unified on trade surveillance practices and systems.
How technology can help manage risk
One of the most common blockers to compliance and regulatory technology (RegTech) is the perception that it will be a hindrance to the business. However, rather than slow down your business, RegTech and service providers are there to protect it – to make sure you’re not running the risk of losing your job or unknowingly engaging in abuse that can cause the firm to suffer financial penalties or reputational impact. Ultimately, good regulatory technology should add value to your firm, rather than be a ‘blocker’ to your operations; it’s not a matter of saying ‘no’ but rather ‘yes; in the right way’.
A trade surveillance system, for example, provides a firm with a centralised exception-based system to manage and escalate trade abuse alerts. Not only does it automate trade surveillance, but these systems are ‘always on’ and work round the clock. As an analyst, it makes spotting the tiny percentage of trade abuse cases among the many false positives that much easier – and these are the cases that really, really matter.
The use of highly configurable parameters and alerts also allows you to align the tech with your trading strategy and for it to evolve with your business. So, as regulations and tech like AI change, it makes it far easier to adapt your trade surveillance processes to match this shift. Crucially, the system acts as a single source of truth, so you have a holistic view of all of the controls that are necessary to manage risk.
Worry about yourself first
While the acceleration of AI brings much uncertainty, in many ways this is no different to other technological evolutions that we’ve lived and worked through in past decades. From HFT to AI, trading environments and regulations have constantly had to adapt and evolve to innovations. You can’t regulate something that isn’t there, but what you can do is manage your processes and controls.
To put it succinctly, if you don’t manage your own risk appropriately, there will likely be consequences for you and your firm. However, if you do manage it well, you’ll help to mitigate that risk as quickly as possible. This approach may sound simple, but simplicity can often be overlooked by firms – especially in the face of the latest technological advances.
