Facebook uses plenty of AI and ML models — but it doesn't seem very worried. Source: Shutterstock

Facebook uses plenty of AI and ML models — but it doesn’t seem very worried. Source: Shutterstock

Why CIOs and CTOs investing in technology should think about model risk

MODEL RISK refers to the risk arising from poor decisions and choices made by artificial intelligence (AI) and machine learning (ML) algorithms and models.

According to a recent collection of articles published by Mayer Brown, the definition of defect or error in algorithms and models itself is subject to a lot of debate — however, model risk has recently become a big concern for many companies lunging themselves into the world of AI and ML.

In the financial services industry, business leaders are seeing regulators get tough with entities whose intelligent models cause the company to take actions that deviate from set norms, standards, or statutes.

Mayer Brown pointed out two instances in its document where the Securities Exchange Commission (SEC) in the US had taken disciplinary action against a quantitative investment adviser and a robo advisor in recent years.

The quantitative investment advisor was found to be at fault because the model in use eliminated one of the risk controls — and the error was concealed from advisory clients. In the case of the robo advisor, the model promised to watch out for ‘wash sales’ (a method of portfolio-based tax optimization) but failed to deliver 31 percent of the time.

On the whole, model risk is something that every business using AI and ML, irrespective of industry or size, needs to watch out for. Facebook, for example, uses plenty of AI and ML models to drive its business, but it doesn’t worry too much because it understands that the benefits of using the technology far outweigh the risks.

In the future, with more companies building applications that use AI and ML models in hiring and appraisal, among other things, care needs to be taken to mitigate if not eliminate model risk.

Are business leaders responsible for model risk?

The short answer to the question is a loud, resounding ‘yes’. Business leaders drive the use of new technologies, including AI and ML and directly benefit from the efficiencies they produce.

As a result, the buck stops with the business leaders when model risks turn into catastrophes.

Truth be told, there’s no reason to fear model risk as it is impossible to postpone innovation or avoid technologies such as AI and ML given the potential they have.

Further, regulators don’t intend to penalize organizations when they use AI and ML. The only reason they take disciplinary action is to ensure that business leaders are not rash when it comes to the use of these technologies and take every precaution to avoid model risk right from the start.

In case something does go wrong with AI- and ML-based models, companies and business leaders often have an opportunity to take corrective action before it is too late — or at least ensure that errors or defects are rectified immediately and don’t repeat mistakes.

Finally, for organizations — especially in the financial services industry — model risk might sometimes be significant enough to damage the organization’s future prospects. In such cases, exploring mitigation solutions or working with insurers to support the business in case of an unfortunate event in the future is a good idea.

Businesses cannot really thrive in the digital-first era and completely remove model risk from its operations. Being aware of it and understanding it can, however, definitely help the business ensure it can balance risks with accelerated AI and ML developments.