Safety and security are top of mind in the wake of a barrage of incidents and developments in 2026, from the revelation that Anthropic’s Mythos Preview was too cyber-capable to release publicly, to the astounding tale of OpenAI agents acting autonomously to launch a cyber attack against Hugging Face. Amid a frontier race with both commercial and geopolitical drives, leading industry figures are increasingly talking about slowing down research to enable greater safety. This panel will discuss cyber and other security risks, the attack-defence balance, misalignment and loss of control scenarios, and opportunities for Australia and likeminded partners to influence global governance, including through the work of bodies like the Australian AI Safety Institute. 

Moderator: Jisoo Kim, Co-Founder and Director, ClearAI

Speakers:
Dr S. Kate Conroy, General Manager, Australian AI Safety Institute
Hamish Hansford, Head of National Security, Department of Home Affairs
Liming Zhu, Research Director, CSIRO