Steering agentic AI in Australian organisations
A new white paper developed by the Governance Institute of Australia, with University of Melbourne’s Centre for AI and Digital Ethics (CAIDE), Mallesons, SEEK, Diligent and EthicAI has brought together industry partners and researchers to explore and advise on the use of AI agents in Australian organisations.
The paper, Governance in the age of agentic AI, sets out the current use of AI agents and lays out advice on key considerations organisations should use when employing agentic AI.
While large language models (LLMs) create specific text-based outputs, agentic AI operates on its own accord, completing a variety of tasks.
Sections included the technical foundations of agentic AI, ethical dilemmas, the legal and regulatory landscape, agentic AI governance and oversight, agentic AI and the board and governing AI agents as a digital workforce.
CAIDE looked at the technical foundations and ethical considerations that come with employing agentic AI.
They highlighted how frameworks designed for humans won’t always work with AI – specific principles may be encoded to the agents, yet how they decide between trade-offs might not be. This raises new questions on ethical responsibility, especially what happens when agents do not act as intended
The paper was launched with a webinar, drawing over 300 governance professionals, with speakers from each organisation talking about their work.
CAIDE Deputy Director, Dr Marc Cheong (School of Computing and Information Systems) spoke about the concepts his team covered in the paper.
“This paper lays out foundations and questions to help guide organisations the use of agentic AI continues to evolve,” Dr Cheong said.
“Agentic AI frequently produces outcomes that are ethically debatable rather than clearly unlawful.
“It’s crucial that as organisations roll out these systems, they consider the ethical implications that come with this often-unpredictable technology.”