Algorithmic Collusion, Herding, and Flash-Crash Risk in Markets: The Unexamined AI Decision Threatens Your Organisation
The risk of algorithmic collusion, herding, and flash-crash in markets is not new. In 2010, the US Congress passed the Dodd-Frank Wall Street Reform and Consumer Protection Act to regulate the financial services industry and prevent future crises. Since then, regulators have implemented various regulations to mitigate these risks, including the introduction of position limits, market data reporting requirements, and stricter capital requirements.
However, a recent study by the International Organization of Securities Commissions found that despite these efforts, high-frequency trading (HFT) firms continue to engage in algorithmic collusion, herding, and flash-crashing. These activities have been shown to be responsible for significant price movements, increased volatility, and potential market crashes.
What does this mean for organisations buying or deploying AI solutions in regulated sectors such as healthcare and financial services? For one, it means that the technology used by these firms must not only perform as promised but also adhere to regulatory requirements. Secondly, the decisions made using these AI systems must be defensible when scrutinised by regulators and courts of law. Finally, organisations are accountable for what their AI does and says.
The development demands immediate attention due to the capability risk it poses – can the technology do what was claimed? The governance risk is equally significant – can the decision be defended when boards or regulators ask questions? And finally, there is liability risk – what is the organisation accountable for when the AI errs?
To mitigate these risks, a prudent organisation should take immediate action by conducting thorough due diligence on the proposed AI solution. This includes examining how the technology works against its real-world operations, people, governance obligations, and legal exposure.
Straven & Co stands ready to help organisations protect themselves from the cost of an unexamined AI decision. The firm independently tests the proposed solution against the client’s real operations, its people, its governance obligations, and its legal and regulatory exposure, delivering a plain verdict: proceed, proceed with conditions, or do not.