The government should avoid “overzealous lawmaking” and lead by example with a “stepped, gradual approach” to regulating artificial intelligence (AI), the Productivity Commission (PC) has advised in an analysis that flags AI’s potential to resolve Australia’s productivity slump.

Aiming to guide the understanding of AI’s productivity benefits, regulatory challenges and implications for data policy, the trilogy of new research papers – collectively entitled Making the Most of the AI Opportunity – concludes, Productivity Commissioner Stephen King said in releasing the papers, that businesses need to think big about AI and “transform core systems and adopt new tools as they emerge.”

As well as investing in AI technologies, the first report notes, businesses must invest in “complementary inputs” such as changes in management, training, and business processes – with such investments likely to be “many times the magnitude” of the technology cost, and likely to take several years to deliver benefits.

The services sector, for one, has notched up early wins and could see AI significantly improve productivity after “adjustment and experimentation” with new technologies – which will initially adopt a ‘co-pilot’ model that assists human decision-making but requires progressively less human oversight over time.

Larger productivity gains will be slower to eventuate, the report notes, with transformation of core business tied to “industry features” including data quality and availability, regulation, investment climate, competition, and “broader features” such as public acceptance.

Governments should lead by example in areas such as AI procurement and use, the report advises, and by demonstrating how data for AI applications can be safely shared, secured, and quality-assured.

The report also warns regulators from getting too carried away with apocalyptic predictions about AI, arguing instead for “proportionate, effective and risk-based” regulations that weigh “potential harms and the risk of harms” against “appropriate real-world counterfactuals”.

“With general purpose technologies in particular,” the report warns, “regulation based on ‘predicted uses’ or ‘speculated harms’ is likely to be overtly broad and harm productivity.”

The advice is a rebuke to often histrionic public debate mired in catastrophising – including fears about “unpalatable” data collection, self-replicating AI robot armies, mass human deaths, Orwellian surveillance, apocalyptic threats, and the implications of OpenAI’s GPT breakthrough, and Meta CEO Mark Zuckerberg’s plans for human-level AI.

Meanwhile, Australian industry has pushed government to move more quickly on clarifying a consistent regulatory environment around AI technology that is progressing at breakneck pace.

The Productivity Commission agrees, arguing that new regulations should only be implemented where risks cannot be managed using existing legislation.

“Formal regulation is only one element of securing safe and ethical AI use,” the report says, noting that “risks of harm are also tempered by social norms, market pressure and coding architecture” – and advising the adoption of a “stepped approach to thinking about AI regulation… that recognises that new technology does not necessarily imply the need for new rules.”

Many of those rules will evolve through Australia’s participation in global rule-setting efforts with the dominance of overseas-developed AI tools meaning that “regulatory approaches in Australia’s key overseas markets will by default act to regulate AI developments and outcomes in Australia.”

“AI will likely highlight gaps in current laws and regulations that we will need to address,” King said, “but pre-empting these gaps with overzealous lawmaking could put local policy out of step with global norms, limiting the potential productivity gains of this new technology.”

Towards a more productive Australia

The report comes as Australian industry wrestles with a protracted slump in “flatlined” productivity that has seen Australians – particularly those employed in the tech sector – working harder but producing less economic value overall, as businesses fall behind in technology usage and employees struggle to recover productivity levels through pandemic-induced remote working.

With up to 90 per cent of Australian jobs set to be affected by AI, its adoption is a shining light in the productivity miasma, promising enough benefits that Treasurer Jim Chalmers recently pushed the PC to “sharpen its focus” on data and digital technologies as a key driver of productivity.

Shortly after warning public servants not to use generative AI (genAI) technology for sensitive work, the Australian government followed the lead of a US Executive Order pushing US government agencies to embrace AI – signing on with Microsoft for a six-month trial of Copilot AI technology that will let public servants access similar technology to that credited with major productivity improvements in other companies.

Westpac, for example, saw coding productivity improve by 46 per cent after introducing genAI while AI security tools have been linked with a 53 per cent reduction in the cost of managing cybersecurity incidents.

Ultimately, the third report notes, successfully harnessing the benefits of AI technologies will require concerted efforts to improve AI data policy – whether through government-level initiatives such as a “comprehensive” national data strategy; “clear and consistent arrangements” to formalise text and data mining (TDM) rights to train AI models; or high-profile commitments around data privacy to build consumer trust.

“The new wave of innovation may finally pull productivity growth out of the shallows,” King said, recommending that “judicious policy interventions and a practical approach to regulation would put the Australian economy in the best position to ride that wave.”