THE AI INTERVIEW
“ When an organisation deploys an autonomous agent, it cannot transfer accountability”
Ivana Bartoletti
Vice President and Global Chief Privacy & AI Governance Officer
Wipro its actual capabilities. This is particularly acute in consumer applications and for children, but it also poses risks in enterprise settings.
Ivana warns:“ When employees trust an AI agent the way they would trust a knowledgeable colleague, they may override their own judgement in ways they would not if the system presented itself as a tool.”
Design, she argues, is part of the solution:“ Systems should be transparent about their nature – not through a disclaimer buried in terms and conditions, but through the interaction itself.
“ They should not simulate emotions they do not have, claim understanding they cannot possess or present relationships that cannot be reciprocal.”
Building a‘ trust stack’ For organisations looking to implement a“ trust stack” – a layered framework of governance controls – Ivana’ s advice is grounded. Before deciding what controls are needed, organisations must first understand what they actually have.
Most, she says, lack a clear picture of what AI systems they are running, who owns them, what decisions those systems are influencing and what data they are using. The starting point is a genuine AI systems audit.
“ The boring answer is always the right answer in governance,” Ivana concludes.“ Know what you have, understand what it does and then decide how to govern it. In that order.”
aimagazine. com 29