Human Control in the Age of AI Agents
Reflections on the UN AI Panel’s new brief: evaluation pressure, external agent state, shutdownability and defence in depth.
Reflections on the UN AI Panel’s new brief: evaluation pressure, external agent state, shutdownability and defence in depth.
Three reflections from ASPI’s AI Masterclass on operational AI capability, speed as a security property, and solve-verify asymmetry.
Reflections on opportunities for French-Australian collaboration in foundational AI, system-level engineering, industrial modelling and sovereign capability.
Reflections on being recognised as a male champion for women in AI, and on how research leaders can create opportunities, offer encouragement and help talent thrive.
The Industry Data Classification Framework helps organisations understand data value and sensitivity, strengthen governance, and enable responsible data and AI innovation.
Reflections from REMIX Summit Sydney on why understanding, rather than judgement or taste, remains an enduring human frontier as AI capability advances.
What happens when AI agents pursue our goals in ways we did not intend? A new piece on alignment, supervisory AI, layered controls and AI sovereignty.
Reflections from two Tech in Gov sessions on sovereign AI capability, harness engineering, technological leverage, and supervisory human oversight for AI agents.
Reflections from the Government Innovation Showcase NSW 2026 on why AI is moving data governance closer to the point of use and into continuous technical controls.
Reflections from a panel on AI in Government during Innovation Month, focusing on capability, context-dependent performance, evidence-based assurance, and continuous governance.




Research Director, CSIRO
Conjoint Professor, CSE UNSW
For other roles, see LinkedIn & Professional activities.
If you’d like to invite me to give a talk, please see here & email liming.zhu@csiro.au