OpenAI and Anthropic have asked Australia to reconsider its refusal to create a copyright exception for training artificial-intelligence systems, proposing a limited regime rather than an unrestricted right to use creative work.
The submissions place a familiar policy conflict inside Australia’s existing copyright framework. Technology companies argue that a carefully drawn text-and-data-mining exception would encourage local infrastructure and model development. Authors, publishers and other rights holders say training should require permission or payment when protected work supplies the raw material.
The government has so far resisted a broad exception. The companies are pressing a parliamentary inquiry to recommend conditions under which training could occur while retaining safeguards for creators. Their argument is also economic: both point to proposed Australian data-centre investments as evidence that copyright policy will influence where AI capacity is built.
The inquiry is expected to report in November. Until Parliament changes the law, the submissions do not alter the rights of copyright owners or provide AI developers with a new defense. They instead define the next legislative contest: whether Australia treats model training as a licensed use, a statutory exception or some combination of the two.
