32 times artificial intelligence got it catastrophically wrong, from chatbots giving truly terrible medical advice to facial recognition software incorrectly flagging people as criminals - up to 78% of failures, fail to trigger any warning or signal.

9 September 2026

ALAN HAYES

 

WHILE the global narrative frequently celebrates automation as an inevitability of modern progress, Australia faces a unique set of geographic, economic, and social vulnerabilities. The rapid integration of generative AI tools threatens to strain the nation’s fragile power grids, compromise the integrity of its world-class education system, exacerbate economic inequality, and dismantle public trust through highly sophisticated cyber threats. As a middle power heavily reliant on resource management, service industries, and social cohesion, Australia stands to lose far more than it gains if the unchecked proliferation of AI continues without strict boundaries.

 

For years, the environmental debate surrounding artificial intelligence focused primarily on the "training" phase—the initial period where a model processes massive datasets to learn language patterns or predictive behaviours. Training large language models (LLMs) requires thousands of specialised graphics processing units (GPUs) running continuously for weeks or months, drawing immense amounts of electricity. However, recent data highlights a far more persistent energy sink: inference.

 

Inference is the phase where a trained model is deployed for everyday use, answering user queries, generating code, or processing images. While a single search query on a traditional search engine uses a negligible amount of power, an AI-powered query consumes up to ten times more electricity due to the complex computational processing required. Because these models are queried billions of times a day globally, inference is now estimated to account for up to 90% of AI’s total lifetime energy consumption.

 

This surging demand has transformed data centres from standard digital warehouses into industrial-scale energy consumers. Globally, data centres consume hundreds of terawatt-hours of electricity annually. In regions with dense data centre clustering, tech infrastructure is consuming an increasingly large share of the total available power supply, threatening to outpace the capacity of local electricity grids. Yet advocates of AI say we have no right to stop innovation in the free market. That’s rubbish. Australia doesn’t allow private corporations to come up with new varieties of cocaine or biological pathogens, just so they can increase the profitability of their bottom line. The government protects the public from certain kinds of innovation. So, why is Albanese besotted with artificial intelligence?

 

The liquid cost of cloud computing

 

While electricity usage is easily quantified, the liquid cost of artificial intelligence remains largely invisible to the end-user. High-performance AI servers generate extraordinary amounts of heat. To prevent hardware degradation and maintain system stability, data centres must rely heavily on cooling systems, many of which depend on evaporating fresh water.

 

The scale of this water consumption is staggering. Global data centres are estimated to utilise trillions of litres of water annually for cooling purposes. A standard interaction with a generative AI platform — amounting to roughly 10 to 50 responses — can effectively "drink" a half-litre bottle of water, depending on the location of the data centre and the season.

 

So, one of the most immediate and tangible challenges AI poses to Australia is its immense physical and environmental footprint. While often conceptualised as a weightless, virtual technology, AI relies on a massive infrastructure of physical data centres. These facilities operate around the clock, requiring immense quantities of electricity to power high-performance graphics processing units (GPUs).

 

Australia's energy sector is currently undergoing a fragile transition from fossil fuels to renewable energy sources. The sudden surge in electricity demand from AI data centres threatens to derail this transition. According to energy market analysts, the power consumption of data centres in Australia is projected to skyrocket, placing immense strain on the National Electricity Market (NEM). This spike in baseline demand risks keeping aging coal-fired power stations online longer than planned or forcing a heavy reliance on gas-fired generation to meet immediate shortfalls. For everyday citizens, this supply-demand imbalance translates directly into higher household electricity bills, as consumers compete with tech giants for grid capacity.

 

Beyond electricity, AI infrastructure demands vast amounts of land and water. Large data centres require significant physical tracts of land, which has increasingly led to zoning conflicts. In regional areas across New South Wales, Queensland, and the Northern Territory, the development of these facilities competes directly with agricultural land and local ecosystems. Furthermore, data centres use millions of litres of water daily for evaporative cooling. In a country uniquely prone to severe droughts and water scarcity, diverting vital water resources to cool server racks has sparked intense pushback from regional communities, who fear for their long-term agricultural viability and local water security. This creates acute localised ecological pressures. In areas already vulnerable to drought, which Australia is, the expansion of AI infrastructure exacerbates water scarcity and places a dangerous burden on local communities.

 

Carbon complications and grid reversals

 

The massive appetite for AI electricity will quickly undermine the decarbonisation targets that have already been set by our government.

 

However, the sheer speed and scale of the AI boom worldwide has outpaced the installation of new renewable energy capacity - a problem that is now evident with the Albanese's prioritisation of fossil fuel to run gigantic data centres in Queensland.

 

Leader of the Australian Greens, Senator Larissa Waters said, “What an extraordinary capitulation on a promise made by the Prime Minister only a month ago that tech giants would need to underwrite renewable energy supply if they are going to drain the grid with power hungry data centres.

 

“The Government themselves admitted that letting power hungry data centres use coal and gas will push power prices up by 13%. Queenslanders would be the hardest hit during this cost of living crisis.

 

“If Labor locks this in, it means that millions of Australians will be paying for AI every time they turn on the lights or air conditioning on."

 

Because AI data centres require a constant, unyielding supply of power — known as "baseload power" — they cannot rely solely on intermittent renewable sources like solar or wind without advanced battery storage. Consequently, utility companies are frequently forced to delay the retirement of fossil-fuel plants or construct new natural gas facilities to meet the sudden deficit. In some overseas regions, the rapid expansion of data centres has actively reversed years of local grid decarbonisation, driving up greenhouse gas emissions and complicating international climate commitments. The environmental toll of AI is, therefore, not just a matter of current footprints, but a structural disruption to the transition toward clean energy.

 

While Anthony Albanese may want to legislate data centre standards, the pushback from jurisdictions like Queensland and the Northern Territory have led to compromises regarding strict mandates on 100% renewable energy sourcing.

 

Australian Greens Leader, Senator Larissa Waters said “If the PM wants to legislate data centre standards, he should start by reverting back to his original commitment to require data centres to be powered by renewable energy.

 

“There needs to be a moratorium on the rollout of data centres until regulation is in place to regulate the impact they have on energy, water use, people and the environment.”

 

The economic and operational strain

 

Beyond environmental degradation, the hidden toll of AI manifests as severe economic and operational stress. Building the physical architecture to support advanced AI requires hundreds of billions of dollars in capital expenditure. This massive financial requirement positions major banking and investment institutions as the ultimate gatekeepers and primary beneficiaries of the boom, pocketing substantial tolls while smaller enterprises bear the brunt of rising operational costs.

 

For enterprises adopting AI, the financial reality rarely aligns with the initial promise of cheap automation. The true cost of implementing scalable AI includes:

 

Infrastructure Tuning: Continuous optimisation of hardware to handle evolving models.

 

Regular Maintenance: Replacing highly stressed, short-lived specialised chips.

 

Security Patching: Defending complex AI pipelines against novel cyber vulnerabilities and data leaks.

 

Furthermore, the localised economic strain impacts everyday consumers. When tech giants demand gigawatts of power to run AI factories, utility companies must upgrade transformers, build transmission lines, and secure extra generation capacity. The billions of dollars required for these grid overhauls are frequently passed down to ordinary citizens through higher monthly utility bills, meaning local populations subsidise the infrastructure of global AI corporations.

 

Workplace disruption and the devaluation of human capital

 

Economically, AI is reshaping the Australian labour market at a pace that regulatory and educational frameworks are struggling to match. Historically, technological revolutions primarily automated manual or routine tasks. The current wave of generative AI, however, directly targets cognitive, creative, and professional occupations—sectors that form the backbone of Australia's service-driven economy.

 

Job displacement and technological unemployment are no longer distant theories but immediate realities for many Australian workers. Roles in administrative support, customer service, legal research, financial analysis, and software development are increasingly being automated or downsized. This rapid shift creates structural unemployment, leaving workers displaced with skills that are difficult to transfer immediately to other sectors. This transition is not seamless; it creates deep anxiety and financial instability for thousands of families.

 

Moreover, the integration of AI tools within the corporate environment has introduced severe workplace friction. The over-reliance on automated metrics to monitor worker productivity — often referred to as algorithmic management — has led to decreased employee morale and increased burnout. In various sectors, the unvetted use of AI to draft communications or make human resource decisions has resulted in disputes, prompting local workplace tribunals and the Fair Work Commission to review and sometimes mandate strict limits on AI-generated workplace communications to protect employee well-being.

 

The creative and cultural industries in Australia face an even more existential threat. Generative AI models are trained by scraping vast quantities of copyrighted data, including the work of Australian artists, musicians, journalists, and authors, largely without their consent, credit, or compensation. This unauthorised appropriation devalues human creativity and undermines the economic viability of the creative sector. The severity of this issue was starkly highlighted when the Australian Recording Industry Association (ARIA) moved to ban fully AI-generated songs from its official music charts, aiming to protect the livelihoods of human artists against an influx of synthetically generated content.

 

The erosion of public trust and democratic integrity

 

Perhaps the most insidious danger, and that is not said lightly, that AI poses to Australia is its potential to undermine social cohesion and the integrity of democratic processes. The rise of sophisticated generative AI has made the creation of highly convincing fake audio, video, and text—collectively known as deepfakes—cheap and accessible to malicious actors.

 

In the political arena, AI-generated misinformation poses a direct threat to democratic stability. During election cycles, local and foreign actors can deploy hyper-targeted disinformation campaigns designed to mislead voters, defame political candidates, and polarise the electorate. The sheer volume of synthetic content floods the media ecosystem, making it increasingly difficult for the public to distinguish between verifiable facts and algorithmic fabrications. When citizens can no longer trust the information they consume, the foundational trust required for a functioning democracy begins to erode.

 

This erosion of trust is reflected in the cultural attitudes of the Australian public. Empirical data consistently shows that Australians are among the most cautious and sceptical people globally regarding the adoption of AI.

 

The Australian Greens have warned that artificial intelligence poses "shocking and imminent risks" to democratic integrity, criticising the federal government for failing to introduce urgent laws to stop AI-driven political deception.

 

The Greens highlighted that under existing laws, it remains legal to circulate unauthorised, hyper-realistic AI-generated video or audio of political leaders — such as the Prime Minister or Opposition Leader — saying things they never actually said, provided it meets basic electoral authorisation rules.

 

Australian Greens Senator David Shoebridge criticised the government for moving too slowly and relying on weak, voluntary safety codes rather than enforceable laws.

 

Surveys conducted by institutions like the University of Queensland and CSIRO reveal that a vast majority of Australians view the risks of AI as a major global priority requiring urgent government intervention. This widespread distrust stems from a well-founded fear of systemic bias in AI models, data privacy breaches, and the lack of corporate accountability from transnational tech companies that operate largely outside the practical reach of domestic Australian law.

 

Profound consequences

 

Artificial intelligence is not a neutral tool; it is a disruptive force that carries profound consequences for the physical and social architecture of Australia. The environmental costs of keeping data centres cool, the economic displacement of professional and creative workers, and the democratic vulnerabilities exposed by synthetic misinformation demonstrate that the unbridled adoption of AI carries severe risks.

 

For Australia to safeguard its future, it cannot rely solely on the self-regulation of the technology sector. It requires robust, proactive legislative frameworks that prioritise environmental sustainability, protect human labour and intellectual property, and defend the public information ecosystem from algorithmic manipulation.

 

AI threatens to exacerbate existing vulnerabilities in the nation's energy grids, labour markets, and social fabric – it is destined to become an environment, economic, social and resource burden upon our nation. But it doesn’t end there! Last month in America, scientists published a study documenting how they used AI to create new kinds of viruses, raising the frightful possibility that the technology could be used to invent dangerous pathogens.

Name (First & Last):

Email:

Submitting Form...

The server encountered an error.

Subscription received.

SUBSCRIBE