We are delighted to be supporting Norton Rose Fulbright's Annual Data Conference with some practical steps that all organisations can take now to manage and mitigate AI risk.

From virtual assistants, to systems that make recommendations for hiring, to automating decisions for credit, to behavioural analysis to better serve a customer segment (to name only a few), AI is being integrated across the value chain in all business domains.

Existing regulators are expanding existing powers and new regulations (such as the proposed EU AI Act) are on the horizon.

Good AI governance is essential to mitigate against material AI risk.

It can be difficult to effectively identify, assess, and mitigate the risks of harms of multiple AI technologies, systems and their uses proportionate to the risk they present:

  • AI raises new ethical issues about how we trust AI systems (which may be opaque or autonomous or may change themselves after they are deployed), the way we use data (where it comes from, if it can be used or re-used), and the potential effects of bias and discrimination when AI makes decisions.
  • Potential risks may arise from training or input data which is biased or incomplete, from AI algorithms or techniques which themselves introduce bias or discrimination, from deployed AI systems or overall business processes which are unfair in their treatment of different categories of people, and from AI algorithms or overall processes which are not transparent or not able to be readily explained or justified.
  • As organisations deploy increasing numbers of AI systems across their businesses, we will find that the output data of one system will be the input or the training data of others. In these eco-systems of AI applications, bias or discrimination may percolate through system after system, causing potential harms well beyond the original source.

Taking early action to inventory your existing landscape, develop processes for capturing important information about AI systems and to learn from data what is driving AI risk can help save time and investment later as more stringent requirements come online.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.