Where should NZ stand on global AI risk? 3 principles to guide a stronger strategy
- Written by: Weekend Times
If some current warnings are to be believed, artificial intelligence may be the human invention that outwits its creators. Understandably, there are now global calls to regulate the technology.
At the current gathering of the United Nations General Assembly, 20 nations (including Australia and Canada) are calling for increased oversight of advanced AI systems. New Zealand – along with the United States and China, which dominate the global AI industry – did not join the initiative.
According to the MIT AI Risk Initiative, incidents of concern are on the rise rapidly. Overall, the global AI industry is raising profound questions about ethics, economy, equity and security.
While there is disagreement between those warning of a potential doomsday scenario and others who take a more benign view of AI’s likely trajectory, investment in the industry is forecast to exceed US$1 trillion this year.
Little wonder the preliminary report of the UN Independent International Scientific Panel on AI found “the gap between rapidly improving capabilities and effective risk management methods may lead to catastrophic outcomes”.
At the same time, the claimed economic benefits of AI often take precedence. Fearing losing the advantage to China, US President Donald Trump has called safety fears a “hoax” and rejected calls for global regulation.
New Zealand Prime Minister Christopher Luxon has argued the opportunities still outweigh the risks. If there are concerns, he says, they are best dealt with in concert with “like-minded friends”.
This optimism is reflected in the government’s “‘light touch approach” to regulation which avoids new laws. (It might also explain New Zealand ranking 32nd out of 45 high-income countries in the Global Index on Responsible AI.)
Conversely, the Labour opposition has promised to establish an Office of AI within the Department of the Prime Minister and Cabinet to coordinate policy across government. It would work with Australia to align regional standards on safety and security.
Neither approach really addresses the kind of existential risk now being discussed globally. To do that, New Zealand could set three goals, each anchored in established international precedent.
Advocate for a precautionary approach
Like most OECD countries, a precautionary principle guides New Zealand policy making: when reputable sources suggest there is risk associated with a lack of action, taking preventive measures against potential harm is justified.
Scientific uncertainty or the lack of immediately obvious negative impacts should warrant an amber light, not a green one. The burden of proof should be on those developing a new technology to demonstrate it is safe through mandatory pre-deployment testing and independent evaluation.
If there is reasonable doubt, pausing development until a technology is shown to be safe should be the default position.
Scientific diplomacy
The UN’s Independent International Scientific Panel on AI lacks the comprehensive governmental buy-in necessary to move international policy forward at pace and scale.
A different model, akin to the Intergovernmental Panel on Climate Change (IPCC), would allow for more coordinated scientific advice, and be larger and more inclusive.
As with climate change, only authoritative, transparent data and shared understanding of the risks around unchecked AI development offer a real chance of building political agreement.
Prioritise the greatest risks
New frameworks can be built in all areas of concern, from ethics to equity. But the highest priority should be those associated with the greatest risk: war and conflict, biological hazards and global financial systems.
Pathways to achieve this already exist – for example, the UN resolution adopted last year addressing the “possible risks of the integration of artificial intelligence into command, control and communications systems of nuclear weapons”.
New Zealand foreign policy could also advocate for the positive use of AI to improve verification of the existing Biological Weapons Convention, as part of the necessary new protocol required to ensure its efficacy.
Similarly, there is an opportunity in November to seek a binding protocol on meaningful human control, accountability and compliance with the laws of war and conventional weapons when the UN meets to review the Convention on Certain Conventional Weapons in November.
International laws and protocols for dealing with cyber-attacks involving AI – which could potentially trigger wars – also need urgent attention.
Finally, at a time when war games and military exercises risk spiralling out of control, New Zealand can encourage confidence-building measures to de-escalate conflicts between geopolitical adversaries.
Here, too, there is emerging precedent: US and Chinese officials only recently proposed a new mechanism for sharing information about serious AI threats. Such developments should be welcomed as a positive first step.
Alexander Gillespie does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.
