Australia to require independent reviews of advanced AI models

What the rules would require

The rules would apply to the most advanced models, those with the highest level of risk. They would be part of the government’s “AI for Australia” program.

Developers of these models would have to:

The last requirement is intended to help the authorities spot early signs of problems, including cases where AI agents begin acting unpredictably. A new regulator, working with the Australian AI Safety Institute, would oversee compliance. Violations would carry heavy fines.

Rules ahead of international negotiations

The bill was announced by Prime Minister Anthony Albanese, Deputy Prime Minister Richard Marles, and ministers Tony Burke and Tim Ayres. They said the government was introducing domestic safety measures ahead of international negotiations.

The initiative aims to protect Australians and national interests, build trust and provide regulatory stability. The government expects this to help attract investment and use AI to benefit the economy and healthcare.

More details will be released after consultations conclude.

Dataist's take: The bill ties access to the Australian market for advanced models to independent oversight and ongoing reporting. The government has yet to disclose the requirements in detail; they will be presented after consultations.