This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility: The ground is shifting beneath the feet of Australian authors and other creatives. A leaked proposal seen by the ABC suggests the government might grant artificial intelligence (AI) companies access to the copyrighted works of Australians under an opt-out system.
There are no firm plans yet, but it appears the government wants to incentivize big AI companies such as OpenAI and Anthropic to bring their business to Australia. Deputy Prime Minister Richard Marles said, "The opportunity for Australia economically is enormous in collaborating with frontier companies." It seems the government also wants to see big AI companies investing in data center development in Australia. What's in play is a choice between the security offered by having local data centers and the long-term interests of Australian writers and other creatives.
While details of the plan are not publicly available, the fact that an opt-out system is being considered suggests a backtrack from the Albanese government's promise that there would be no text and data mining exception to copyright for AI companies. Under such an exception, Australian copyright holders would be compensated for the use of their works in AI training. However, Minister for Industry and Innovation Tim Ayres has denied a reduction in copyright protection.
An opt-out system doesn't gel with Australia's current copyright laws. Under the Copyright Act, the owner of the copyright in a piece of writing or photo has the right to exploit those rights as they choose. Forcing copyright owners to opt out of a given system—such as their works being included in training materials for AI models—impedes their rights because it forces them to take action to protect those rights.
Our copyright law dictates that the burden should run the other way: AI companies should be asking for permission. The problem with an opt-out system is that by the time you get around to actually opting out, your materials have likely been inside the neural network for a while. Once materials are in the model, the AI system can't "unlearn" them.
This is where the heart of this dispute lies. The problem is not just the taking of individual copyright-protected works. It's the ever-increasing ability of AI systems to produce quality materials as a result of training on copyright-protected works.
Extract — continue reading at the source.
