ABSTRACT
Artificial intelligence (AI) has already proven itself to be transformational and pervasive: transformational as it is radically changing the industry it impacts and pervasive because it is impacting everything from clinical trials and drug discovery to creative content generation. This is a moment in technology development akin to the arrival of computers.
Therefore, at such a defining moment, it is important to ensure that there are mechanisms to support AI innovation so that it can work with local regulation, be evaluated in the economic ecosystem and be able to defend itself in a legal liability incident. (This is all the more important when the AI system is used by the state/government, is funded by taxpayers and has to win the trust of residents of a jurisdiction, so that higher efficiencies can be achieved in delivery of civic services.) A lot of apprehensive conversation is around AI’s potential to perpetuate and magnify current inadequacies and biases in human-controlled decisions and forecasting and AI’s potential to cause harm in the absence of meticulous governance, safety guiderails and oversight.
This chapter proposes a risk assessment methodology, which creates a complete risk profile of an AI system based on the ecosystem in which it intends to operate, data and model governance considerations. Authors further propose a quantification approach for this risk profile and suggest pathways for its use by the regulator, its use to generate funding/investment, interacting with insurance industry and, most importantly, the role a risk profile will play in incident management and liability determination.
