AI friend or foe
What does the future of AI legislation look like?
The debate surrounding AI has already moved from what it can do to what we should allow it to do. So just what does the future of AI legislation look like? The experts weigh in.
Featured experts
When the hardware retail giant Bunnings was called out for its use of facial recognition technology to scan customers on entry to its stores – for the laudable purpose of reducing theft and increasing safety – the effect was felt across the world. It was labelled “a privacy breach on a massive scale” – and the clearest indication yet that the discussion has moved from what AI has the potential to do, to what we should allow it to do.
From intellectual property to criminal responsibility (and, of course, the biggie – the rule of law), choosing frameworks to think about and regulate AI will be key to its success. But how can we ensure AI is a benefit and not a menace, and what should we really be thinking about when we think about AI?
“The 2024 AI Act in the European Union put a stake in the ground,” says Professor Joanna Batstone, inaugural Director of the Monash AI Institute. “It was risky to make the first move, but it prompted other governments and jurisdictions to consider their own positions.”
Risk-based approach
The EU approach is risk-based, so many of the conversations are about categorising different styles of use-cases in terms of high, medium and low-risk. According to Batstone: “Some of the high-risk scenarios that are often discussed include things such as adverse impacts to an individual’s human rights as it relates to a human rights law in a particular country or jurisdiction; adverse impacts to an individual’s physical or mental health or safety; adverse impacts to the economy to society; or, particularly here in Australia, effects on groups of individual or collective rights of cultural groups or Indigenous communities.
“So there’s a tiering and categorisation of things that are viewed as high-risk, and while I understand that you can’t regulate everything, I’d suggest it’s a bit of a slippery slope to not worry about the low-risk ones as much.”
The search for a global standard
Like the EU’s General Data Protection Regulation (GDPR) in 2018, the EU AI Act could become a global standard, determining to what extent AI has a positive rather than a negative effect on our lives, wherever we may be.
But, says Barb Hyman, Founder and CEO of Sapia.ai, AI is too broad to regulate at a macro level, acknowledging that ambiguity increases risk for everyone.
“Although the EU legislation made life hard for tech companies in many respects, it did create certainty around what your obligations are. Here in Australia, the federal government says we should be AI-first and really embrace it as a nation, but the number of standards and principles it has come up with doesn’t really do very much – other than create uncertainty and ambiguity for both sides.”
The focus is not about regulating a particular technology, but instead regulating the human behaviour relating to that technology."
Dr Ridoan Karim, Lecturer in Business Law, and Deputy Director of the Centre for Commercial Law and Regulatory Studies (CLARS) – Malaysia Hub, at Monash University, Malaysia, says that the aim needs to be to make legislation technology neutral, so that the focus is not about regulating a particular technology, but instead regulating the human behaviour relating to that technology.
And for that to happen, he says, we need a global consensus on how we want AI to benefit society.
“That might be through international instruments, or it could be part of a consensus, similar to what was developed around nuclear technology – we see the devastating nature of the nuclear technology, and there are treaties in place to say that we will only use this technology for the betterment of the human population, not for the destructive nature.”
New regulation, new thinking
To achieve the true potential of AI, Karim suggests, we need to establish some basic principles about just what the point of AI is.
“When we talk about regulating AI, it doesn’t necessarily need to be regulated with laws,” he says. “New technology requires new thinking, so we should establish a consensus on general basic principles around the ethical and moral practices of AI, and then leave individual societies to tailor their own futures.
“Industry and corporations say they want to be regulated. The thing that these corporations are afraid of is not having equal opportunity in every region of the world, or having too much fragmented regulations. They want to do business, and they understand that to do business, you need regulations in place.”
Professor Chris Marsden, Professor of Artificial Intelligence, Technology and Law at Monash, agrees. He’s interested in what he calls a “digital duty of care”, and cites the recently-approved social media ban on under-16s.
“When people think of AI, they think of it in terms of the generation of harmful content,” says Marsden. “In fact, it’s used more by the social media platform to try and save us from human or machine-generated harmful, otherwise misleading or spam content.”
But it’s also about the potential for digital duty of care, which is being discussed at the moment by policymakers, that puts more duties on platforms to make sure they use AI to flush out that harmful contact. “That’s a much more interesting idea, and follows up the European approach towards what they call maintaining human dignity.”
Factoring in the individual
Individual responsibility is another key factor in how the technology will develop, and Dr Paul Burgess, senior lecturer and academic in law, points to the example of another developing resource from 20 years ago – Wikipedia.
“Alongside any regulation, there needs to be an emphasis on our own use and understanding the limitations of the technology.
“When Wikipedia first appeared, there was still a reluctance to assume its content was necessarily correct. Now, we’re more confident in our use of Wikipedia, and we have an appreciation of the way that its content is created. Am I going to use Wikipedia to find the most effective technique to perform open heart surgery? No – and I really hope my heart surgeon isn’t either.
“We didn’t need to regulate whether Wikipedia should exist or the content it should have on there. But we did need to learn how to use the resource a little bit more effectively and efficiently, and understand there are levels of just how accurate information is.”
A need for trust
Trust, then, is important, something Associate Professor Campbell Wilson, co-founder of the Artificial Intelligence for Law Enforcement and Community Safety (AiLECS) Centre, is keenly aware of.
“Open-source models are a particular challenge for regulation,” he says.
“To take child protection as an example, we’re particularly concerned with the way generative AI is being used around child sexual abuse material at the moment. The tech platforms do put in a lot of guardrails around the generative AI algorithms they release, but the content that’s hosted on open-source platforms are subject to very little control, and it’s really hard to regulate them. Transparency is crucial.”
I think we have to take on a responsibility to focus on AI systems that are built to operate safely and ethically.”
Batstone emphasises the need for any regulation to be mindful of safety. “I think we have to take on a responsibility to focus on AI systems that are built to operate safely and ethically.”
But what does AI safety look like? It’s an active discussion in Australia and globally to understand what safe, responsible and ethical mean in the context of AI technology, “so that we have collectively a global set of guardrails being defined, that enable collectively the organisations to chart a path forward”, says Batstone.
So is AI friend or foe? With developing technologies seemingly inserted into every aspect of life, and new developments around every corner, the opportunities and risks come thick and fast, and any regulation needs to not only keep pace, but ensure it encourages, not restricts.
“There are plenty of things in development, but Australia doesn’t yet have any kind of general law about AI,” says Batstone. “But it’s time we did, because clearly this is too profoundly important for us not to use it.”
