Last week, representatives from Amazon, Google, and Microsoft appeared at a public hearing before the Select Committee inquiring into the Adoption of AI in Australia. While we were pleased to see Government representatives putting creator concerns to these tech companies, the testimony of Amazon and Google was alarming – not only because important questions about the ingestion of copyright material remain unanswered, but also because their arguments seemed to rest on assumption of entitlement to copyright works to optimise their products, without consent from or compensation to copyright owners.

Amazon and Google dodged straightforward questions about the copyright works ingested to train their AI models. Google was specifically asked whether the text in Google Books was used to train large language models to which representative, Ms Longcroft, Director, Government Affairs and Public Policy, said, “our models are trained on publicly available information that are crawled on the open web.” Amazon was also asked directly whether the enormous volume of filmed, written and recorded work of Australian creators through Prime Video, Kindle, and Audible is used to train AI models. Mr Levey, Head of Public Policy, Australia and New Zealand at Amazon responded, “I am not familiar with the answer to that question Senator.”

Amazon did agree to take the question on notice but it is galling when responses to fundamental questions – which have been the subject of a global outcry from creators for over a year – are answered with obfuscation.

Senator Tony Sheldon, Chair of the Select Committee, said in the hearing, “It defies belief that you’re unable to tell us whether Amazon has used the work of Australian creatives across Audible, Kindle and Prime Video to train your AI models. Your answers provided no comfort to people worried about getting ripped off as a result of their life’s work. If I was a writer or voice actor listening to this I would be terrified.”

Throughout the hearing, the argument run by both Amazon and Google was that it was necessary to include an enormous amount of copyright works and data in their training datasets, to ensure their generative AI models would be less biased and more effective. Yet the option of licencing that work was simply never mentioned. Instead, these companies seek to change the law to grant themselves permission to use the work of creators without consent, credit, or remuneration.

The ASA will continue representing authors’ and illustrators’ concerns to Government and working with other industry bodies to urge the Government to take action to protect the livelihood of creators.

In other local news, music rights management organisation APRA-AMCOS this week released the results of a report into AI in the music sector exploring how music creators are using AI and the potential impact of AI on music creation and value. Particularly concerning were the estimated damages to the industry totalling half a billion dollars, and the concerns 82% of respondents share that AI use in music could lead to them no longer being able to make a living from their work. While this report specifically addresses the music sector, we expect there to be similarities to the literature sector. Macquarie University will be conducting research into the use and impacts of AI upon authors and illustrators in Australia. Stay tuned to our newsletter for information about how you can contribute.

In international AI news, we were buoyed to see that a US Court has agreed to advance all copyright infringement and trademark claims against AI art generator, Stable Diffusion, in what represents the first ‘win’ for artists in this case. The lawsuit was brought by a group of artists in early 2023, who claim that Stability AI infringed their copyright when it used their work to train Stable Diffusion without authorisation or compensation.

We will keep members updated of any new developments in the overseas lawsuits as they arise.