In an interim report released overnight, Harnessing data and digital technology, the Productivity Commission has floated a text and data mining exception for the Australian Copyright Act.

This would make it legal to train artificial intelligence large language models, such as ChatGPT, on copyrighted Australian work. AI training would be added to the list of “fair dealing” exceptions already existing in the Copyright Act.

Why? The Productivity Commission estimates a potential A$116 billion over ten years flowing into the Australian economy, thanks to AI.

Of course, this comes after large language models have already “trained” on masses of Australian copyright material, breaching copyright law. In March, many Australian authors were outraged to find their works included in a dataset of pirated books used by Meta to train their AI systems (including books by former prime ministers John Howard and Julia Gillard).

Writers, publishers and their industry bodies oppose any such exception – which would “preference the interests of multinational technology companies at the expense of our own creative industries”, according to the Copyright Agency. And this isn’t the first time the Productivity Commission has proposed changes that would harm Australian publishers.

Sophie Cunningham, a........

© The Conversation