Australia is stuck in the “worst of possible worlds” when it comes to artificial intelligence and the use of local data, according to Productivity Commission chair Danielle Wood, arguing that reforms to copyright law could help Australian companies compete globally and strengthen the local AI industry.

Wood addressed the National Press Club in Canberra on Monday ahead of the federal government’s productivity roundtable, which kicks off in Canberra today.

It comes after the Productivity Commission released a series of interim reports including recommendations for promoting productivity across five pillars, including data and digital, and skills and the workforce.

One of the proposals from these reports that has received the most attention is for a “fair dealing” exception that would allow Australian companies to use the copyrighted material of Australians without permission to train their artificial intelligence large language models (LLMs).

While not making a recommendation on the topic, the Productivity Commission is now “seeking feedback about the likely effects of a text and data mining exception on the AI market, the creative sector and productivity in general”.

The proposal quickly received widespread backlash from Australian creatives and copyright holders, and online privacy and digital rights organisations.

At the Press Club, Wood said that this is merely one of several options being considered, but said that some of the criticisms have missed the point.

“I think the pushback misunderstands some of the context of that particular discussion,” Wood said on Monday.

“At the moment we are in what I would call the worst of possible worlds. Models trained overseas are using Australian data – Australian creative content – and they are able to do that because it depends on the laws of the nation where that model is hosted.

“Nothing in Australian copyright laws can make a difference to that.”

But reforms to local copyright laws could help local tech companies compete with the world, Wood said.

Wood said AI models hosted overseas are already being trained on the work of Australian creatives. Photo: Shutterstock

“The world we’re in creates uncertainty for people who might want to build models in Australia,” she said.

“There’s a lack of clarity and potentially very high transaction costs if they want to use Australian content.

“That’s the question we were focused on: how can we support an Australian AI industry?”

No need for AI-specific laws

In the speech, Wood also reiterated the Productivity Commission’s stance that the government should abandon any plans for AI-specific regulations or laws, saying that this could stifle growth and prevent Australia from capitalising on the opportunities on offer by the technology.

She said that AI has the potential to be a transformative technology and could add more than four per cent to labour productivity over the next decade, equating to an extra $116 billion in economic activity or an income boost of $4,300 for each Australian per year over that time.

“A growth mindset means that we must not regulate our way out of this opportunity,” Wood said.

“While managing the risks is important, we do not think that a new and overarching regulatory framework for AI is the way to go.

“That’s because the risks posed by AI are mainly existing risks.

“AI may make it cheaper, easier and faster for bad actors to create harms, but most of these harms – from product safety, to discrimination, to fraud – are already covered by regulatory frameworks.”

Instead, the federal government should quickly continue its work to plug gaps in existing frameworks when it comes to AI, Wood said.

“This work should be done urgently and comprehensively and in preference to broad-based technology regulation, which has been shown overseas to lead to reduced innovation, digital jobs and investment,” she said.

A report by the Productivity Commission earlier this month said that AI-specific laws should be a “last resort”.

The federal government has reportedly moved away from introducing an AI Act in the wake of the Productivity Commission's comments.

AI and education

There also needs to be a push for all students and teachers to have access to high-quality learning technology and materials, including AI.

“We have a patchwork of approaches to edtech and AI in the classroom across our school systems,” Wood said.

“And while some schools have high-quality curriculum materials that free up teacher time and ensure the quality of student lessons, not all do.

“The result is something like a lesson lottery.”