Agentic AIs can act both remarkably and unpredictably. Imagine setting up a few autonomous AI agents to perform simple tasks, only to watch them behave in unexpected ways - exposing privacy, affecting user rights, or refusing to respond at all. These are not mere “hallucinations” but agents going rogue. In this interactive session, we’ll explore why that happens and how a simple, human-centred approach can prevent it.

Through live demonstrations with everyday robots and computers, Dr Vishal Sharma and his team (Fannya R. Sandjaja and Gyan Wickremasinghe) will invite you to participate, step in as the crucial human in the loop, and discover how thoughtful coding practices keep AI safe, reliable, and under control.