GenAI – the technology behind ChatGPT – is as transformative and disruptive as the internet. While the internet revolutionised distribution and access to content, genAI transforms content creation. As genAI relies upon large-scale unconsented and uncompensated use of training data, this poses fundamental challenges for laws establishing rights in content, especially copyright.
In Australia, attention has focused on the uncompensated use of news content as training data, but views are polarised. On the one hand, Google has argued for more flexibility in copyright law, supporting a new exception for text and data mining. On the other hand, both Nine and News Corp have proposed payments for using news content by genAI, upon analogy with the news media bargaining code.
The issues facing news content are a subset of the general issues facing copyright. Copyright is designed to protect creative content by preventing free riding by uncompensated copying. While genAI uses copyright content, however, the sheer scale of the training data means that outputs may not be infringing copies. This leaves aside potential infringements by copying input data, with differences in how this may be treated under national laws.
Currently, there are several cases alleging copyright infringement by genAI systems making their way through the US courts. Meanwhile, suggestions for addressing the problem of uncompensated use have included a compulsory licence, which would allow potentially infringing uses provided payments are made to rights holders. The problem with this, however, is that the enormous scale of the data training sets would make it difficult or impossible to identify and pay rights holders.
While it is likely to take time to develop solutions to the copyright problem, this need not prevent consideration of the particular issues facing news production.
In our view, there are at least two features that distinguish news from other copyright content. First, as recent history has illustrated, accurate news reporting is vital to functioning democracies. Secondly, as news content producers are a more homogeneous group than copyright holders in general, it is easier to identify and pay rights holders. We, therefore, think that, at least as an interim measure, there is a good case for establishing a mechanism for compensating news providers for the use of their content in genAI, in the form of either an extension to the news media bargaining code or a new compulsory licence.
– Evana Wright and David Lindsay, UTS Law
This was featured in our Centre's fortnightly newsletter of 11 August - read it in full here and/or subscribe.