Australia should adopt US-style copyright law to allow artificial intelligence to use vast amounts of creative content, or risk damaging investment in the creative industry, according to Atlassian co-founder, Scott Farquhar.
Farquhar, CEO of the Tech Council of Australia, told ABC’s 7.30 program that using AI to mine or search through data is currently probably against Australian law, and that this undermines investment in AI companies.
This is because Australia’s copyright law lacks the “fair use” exemptions found in US law, which permit the unrestricted use of copyrighted material for specific purposes such as criticism, comment, news reporting, teaching, scholarship, or research.
But Farquhar’s claim overlooks the fact that “fair use” is not settled law in the US and could have disastrous consequences for the creative industries.
Companies such as Atlassian, Google, and Meta want a text and data mining exception added to copyright law to enable AI technology to train on all human works without payment.
Farquhar suggests that AI usage of copyrighted material is not theft unless the AI is used to “copy an artist directly,” such as by creating a song in their style.
However, US law does consider whether the AI’s output is “transformative,” or fundamentally new and different from the original work, when evaluating “fair use.” Farquhar believes that the benefits of large language models outweigh any issues related to their training data, which may include other people’s original work.
The US Copyright Office has stated that AI training on copyrighted works without permission poses a risk of financial harm to the market for or value of those works. The office has, however, suggested that AI development can continue without undermining intellectual property rights if voluntary licensing agreements are put in place.
Many industries could be adversely affected by the use of AI. For example, AI-powered summaries of news articles have already reduced clicks through to the full stories online. It ignores the fact that fair use in relation to AI is not settled law in the US and that courts are actively debating the issue.
