IOSCO: AI & Machine learning report – key points

As AI technology will be a major driving force behind successful business strategies in the financial services market in the next years, IOSCO’s Guidance provides financial institutions with general indications to govern it with great care and scrupulousness.

Governance and oversight

The guidance recommends that regulators should mandate the implementation of robust governance procedures concerning AI and ML, requiring firms to designate members of senior management with the relevant skillset and knowledge to adequately oversee this technology. The guidance states that proportionality should underpin corporate governance procedures concerning AI and ML, taking into consideration the impact on market and client outcomes as well as the nature and complexity of the activity being performed.

Algorithm development, testing, and ongoing monitoring

The guidance states that regulators should require firms to design and implement procedures for development, testing, implementation and monitoring of AI and ML systems to ensure that they operate in a way that complies with the firm’s regulatory obligations. Adequate testing should be carried out in both stressed and unstressed market conditions.

Data quality and bias

The performance of AI and ML is largely dependent on the quality of the dataset(s) used when building the model. The quality of datasets can have a material effect on the outcome of AI and ML applications. The guidance states that regulators should consider a requirement that firms implement appropriate data quality controls to ensure that the use and development of the AI and ML deployment does not jeopardise those firms or their customers.

Transparency and explainability

The successful use and adoption of AI and ML requires that algorithms not only be accurate but also be understood by firms, firms' clients and regulators. Understandability helps take correct measures to mitigate the possible risks arising from the AI systems. The guidance states that regulators should consider requiring firms to disclose pertinent information to customers and clients around their use of AI and ML, especially those that affect clients’ outcomes. Regulators should also consider what disclosure they may require enabling them to oversee those firms appropriately.

Outsourcing

A lack of adequate in-house skills, experience and expertise to sufficiently oversee AI and ML could lead to over-reliance on third parties providing outsourced AI services. The IOSCO guidance states that regulators should consider requiring firms to ensure that compliance and risk management units as well as senior management are able to understand and monitor the provision of AI and ML services by third-parties. In this regard, clear lines of accountability can be established and maintained.

Ethical concerns

Ethical concerns may arise if AI / ML models develop certain social biases that could potentially lead to the production of undesirable outcomes. Biases can exist within the data sets utilised. The guidance states that firms should consider criteria to minimise the risk of discrimination results.

Our FinTech experts are ready to provide all the entities in the financial industry value chain with legal assistance in setting a sound Governance of the AI systems, capable to address the regulatory and contractual risks associated with their use, to mitigate unintended and unwanted effects, consistently with the IOSCO Guidance.

If you would like to read Bird & Bird’s previous alerts, please check out our FinTech In Focus webpage here.