What gaps need to be addressed for AI regulation: TCA CEO

As the federal government continues to take submissions over the safe and responsible use of AI some tech leaders in the industry have put together a wish list of what they would like addressed.

Kate Pounder, CEO of the Tech Council of Australia (TCA) spoke to Digital Nation about what gaps need to be filled in the upcoming regulation.

With the current laws such as the privacy act, copyright act and Australian consumer law, Australia isn’t starting from scratch in terms of building regulation, Pounder noted. 

“It is an important thing to understand, it is firstly reminding people who are developing their technologies, but people using them that these frameworks do exist, what their rights are and what their responsibilities are,” she said.

Pounder said where some of the gaps will be is where the technology has created use cases that were not really contemplated by those laws.

“That is where privacy reform is going to be key in Australia. We haven't done a proper update of our privacy act in years. It is not as modern, it's not consistent with global practice at the moment,” she said.

“AI relies on data to be accurate, to be fair, and to be transparent. If you don't get that underlying data regime and privacy regime right it's harder to expect these technologies will have the outcomes that they want.”

Another central regulation will be around intellectual property (IP) and copyright. Pounder said this is especially needed around generative AI.

“Any form of AI will depend on the quality of the data sets that it's trained on. That's an interesting area of law where it was premised on this idea of scarcity and the need to incentivise production. Whereas AI, is almost totally flipping the economics of that equation of creativity,” she explained.

“There is a real argument that if you care about the fairness, the transparency and the representativeness of AI, you don't want to be limiting or making it impossible for diverse training data to be used to develop these training data sets and for AI to learn on.

“It has negative impacts in some cases, IP and copyright are going to be a central question to work through.”