Introduction
This is the second of 2 key updates since the publication of the International AI safety report in January 2025. It was written by independent experts and aims to provide scientific information to support informed policymaking.
Australia’s representative on the expert advisory panel was Dr Liming Zhu, Research Director, Data61, CSIRO. Australian civil society and industry contributors included:
- the Gradient Institute
- Old Ways New
- Harmony Intelligence
- Good Ancestors Policy.
The International AI safety report aims to synthesise scientific evidence to support informed policymaking. It does not make specific policy recommendations.
This update presents the following key findings:
- 12 companies have published their own frontier AI safety frameworks in 2025. However, implementation can be mixed, and these commitments are difficult to verify externally.
- While safeguards preventing the misuse of AI models have improved, they remain fragile against sophisticated attacks and complex tasks.
These results reflect an increasing emphasis on managing the risks of AI technologies and highlight the need for further research and practical solutions in this area internationally.