Australia Advances AI Safety Science at International Network Meeting
In July, Australia participated in the third directors-level meeting of the International Network of AI Safety Institutes held in Vancouver. This gathering brought together key representatives to push forward scientific research focused on artificial intelligence (AI) safety.
The network serves as a platform for international collaboration on AI safety science, building on commitments made in the Seoul statement of intent from the AI Seoul Summit on 21 May 2024.
Global Collaboration on AI Safety
Australia joined members from Canada, the European Commission, France, Japan, the Republic of Korea, Singapore, the United Kingdom, and the United States. The meeting took place alongside the 42nd International Conference on Machine Learning (ICML) in 2025, a leading event dedicated to machine learning research.
Research and Testing Initiatives
The network published key papers to guide discussions at ICML. Australia and Canada co-led a research agenda focused on managing risks associated with AI-generated content. Additionally, Singapore and the United Kingdom co-published findings from the network’s third joint testing exercise.
This testing exercise aimed to enhance the scientific methods for evaluating AI agents and to develop shared best practices. It was divided into two strands:
- Leakage of sensitive information and fraud – led by Singapore
- Cybersecurity – led by the United Kingdom
Australia contributed technical expertise to both strands, with input from organisations including CSIRO’s Data61, the Gradient Institute, Mileva Security Labs, and UNSW’s AI Institute.
Ongoing Commitment to AI Safety
The Australian department involved will continue supporting the network’s mission to improve AI safety. This includes helping governments and communities understand potential risks posed by advanced AI systems and proposing practical solutions to reduce harm.
For professionals interested in enhancing their AI knowledge, Complete AI Training offers a range of up-to-date courses covering AI safety and related topics.
Your membership also unlocks:
