What was announced
According to The Decoder (DE), the government under Prime Minister Anthony Albanese has announced a law that introduces binding safety rules for frontier AI companies that want to operate in Australia, for the first time. The rules target the technically most advanced models with the highest risks and are part of the agenda "AI in Australia's Interests".
The planned obligations
Affected companies would have to open their models to independent testing, ensure ongoing safety checks, publish their safety frameworks and report security incidents to the government immediately. According to the report, the reporting duty is meant, among other things, to help give early warnings about threats from AI agents that go out of control. A new oversight body is to monitor compliance together with the Australian AI Safety Institute and could impose steep penalties for violations.
Open questions
Further details are to follow only after consultations conclude. The source does not yet name an entry-into-force date or concrete penalty amounts. According to the report, the government wants to set the domestic guardrails before international negotiations.
Context
If the law comes as announced, providers of frontier models operating in Australia could face additional testing and reporting duties. Whether and how this affects companies in Germany depends on details that are still pending.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
