Risks of Artificial Intelligence in Finance

Artificial Intelligence techniques are increasingly deployed in finance, in areas such as algorithmic trading, blockchain-based finance, and asset management, enabled by the abundance of available data and affordable computing power. Machine learning models are using deep learning algorithms based on vast amounts of data, to learn and improve predictability and performance automatically, without explicitly being programmed by humans.

The advancement of AI in finance and other areas of business is expected to increase competition, and induce competitive advantage for businesses, by improving their efficiency, cost reduction, productivity enhancement, and providing a better quality of services and products to consumers.

However, AI applications in finance may create or facilitate financial and non-financial risks, and give rise to potential financial consumer and investor protection considerations. The lack of explainability of AI models could give rise to systemic risk in the markets, and could create possible incompatibilities with existing financial policymaking. While many of the potential risks associated with AI in finance are not unique to this innovation, the use of such techniques could amplify the vulnerabilities given the complexity of the techniques employed, their dynamic adaptability, and their level of autonomy.

The propagation of AI in financial systems

Artificial intelligence systems are machine-based systems with varying levels of autonomy that can, for a given set of human-defined objectives, make predictions, recommendations, or decisions.

These systems are increasingly using copious amounts of data, big data. Data feeds machine learning models which use the data to learn and improve predictability and performance automatically through deep learning algorithms, without being explicitly programmed by humans.

The COVID-19 crisis has accelerated the digitalization trend that has already been observed before the pandemic. Global spending in the AI market has doubled in the last 2 years and is forecast to double over again in the next two years. Growing AI adoption in finance, in areas such as asset management, algorithmic trading, credit underwriting, or blockchain-based financial services, is enabled by the abundance of available data and by increased, and more affordable, computing capacity.

Effects of AI on trading

AI applications are applied in asset management and the buy-side of the market for asset allocation and stock selection. ML models have a profound ability to identify signals and capture underlying relationships in big data, where financial quantitative analysis or human intuition fails to do so.

Quantitative analysis relies on graphs and mathematical data, human brainpower and intuition can analyze assets based on emotions, and feelings. A successful investor uses both to his advantage.

But an AI system can use the two methods in tandem, and also with substantially greater number of data, analyzed in split seconds. This puts the AI system, and the AI-enhanced investor at an unfair advantage against traditional investors.

Risks of algorithmic trading based on AI models

Similar to non-AI models and algorithms, the use of the same ML models by a large number of finance practitioners could potentially prompt herding behavior, which in turn may raise risks for liquidity and stability of the system, particularly in times of stress.

Although AI trading can increase liquidity during normal times, it can also lead to convergence and by consequence illiquidity during times of stress and flash crashes. Market volatility could increase through large sales or purchases executed simultaneously, giving rise to new sources of vulnerabilities. The convergence of trading strategies creates the risk of self-reinforcing feedback loops that can, in turn, trigger sharp price moves.

Such convergence also increases the risk of cyber-attacks, as it becomes easier for cybercriminals to influence agents acting in the same way. AI techniques could exacerbate illegal practices in trading aiming to manipulate the markets and make it more difficult for supervisors to identify such practices if collusion among machines is in place.

This is enabled due to the dynamic adaptive capacity of self-learning and deep learning AI models, as they can recognize mutual interdependencies and adapt to the behavior and actions of other market participants or other AI models, possibly reaching a catastrophic outcome without any human intervention.