Canberra is ‘consuming the Kool-Aid’ on AI. Where will it take us?

There was a revealing minute last month when Communications Minister Anika Wells appeared on the ABC’s Experts to talk about the federal government’s most current task of care arrangements for social networks. Inquired about whether the federal government desired AI advancement to decrease, she let slip an amazing aside.

“AGI is here, isn’t it?”, Wells statedmatter-of-fact, describing “synthetic basic intelligence”.

The declaration was not extensively reported. An ABC article even altered Wells’ quote to get rid of the “G”, altering her remark to merely, “AI is here, isn’t it?”

AGI– a still-hypothetical advancement of AI that might find out, factor and adjust throughout a broad variety of jobs with restricted human oversight– is the nub of the concept behind Donald Trump’s favoured term, “extremely intelligence”. Its presence is likewise strong subject of dispute amongst cognitive researcherspsychologists and innovation scientistssince AGI’s introduction would declare an extensive ethical, social and recession for mankind.

Maybe Wells was echoing Nvidia employer Jensen Huang, who last month declared “AGI has actually gotten here”. Regardless, her quip informs us something essential about an ambient environment of AI adoption pervading the Canberra policy discourse, one greatly affected by the trillion-dollar buzz maker of Silicon Valley.

Due to the fact that even if “AGI” isn’t “here”, the federal government makes sure imitating it is.

‘ Drinking the Kool-Aid’

Anthony Albanese most likely wasn’t expecting AI to be among the essential policy obstacles of his federal government. The quick climax of the market in current years has actually made that unavoidable. When an extremely specific niche subject shelled out to the boffins and mid-level bureaucrats, AI policy has actually relocated to the centre of political conversation.

Internal federal government questions culminated in the release of the National AI Plan last December. Its framing is finest summarized by the title of its very first chapter, “Capturing the chances”, which breezily recommends “Australia’s success in recording the chances of AI depends upon equipping our individuals to get the very best out of AI in their tasks, neighborhoods and their individual lives”. The file canvasses a collective effort to “draw in financial investment” and just navigates to emerging dangers from brand-new innovations towards completion.

Optimism about AI continued in early 2026, increased by a check out to Canberra from Anthropic’s Dario Amodei. Providing himself as the smiling face of ethical AI research studyAmodei was at the time indulging in the radiance of his business’s effective Claude design. With an entourage of flunkies and geeks, the CEO protected conferences with Albanese, Treasurer Jim Chalmers and Assistant Minister for Science, Technology and the Digital Economy Andrew Charlton, while likewise hosting a parliamentary online forum that ended up being a hot ticket for Labor backbenchers and civil service leading brass.

A senior operative in a Labor minister’s workplace explained Amodei’s reception to Crikey as “consuming the Kool-Aid”.

The CEO’s rock-star treatment partially comes from the quick uptake of chatbots and big language designs (LLMs) by Canberra elites themselves. There can be no rejecting the expansion of these tools in the senior civil service and amongst numerous policy, market and scholastic elites. ChatGPT’s record run from early 2023 is extensively acknowledged as the fastest-adopted customer application in history

This AI buzz, which has actually assisted mask the naked avarice of Silicon Valley, has likewise contaminated big parts of Australia’s supervisory and expert elite. The well-informed chattiness of Claude or ChatGPT appears especially appealing to a particular class of senior executive, university vice-chancellor and department secretary. Looking like an abnormally intense and effective research study assistant, AI chatbots are ever all set with a handy summary, a proof base, or a pre-formatted table with all the details set out so.

Such are the disordering impacts of this mania that even senior union leaders like the ACTU’s Joseph Mitchell appear to have actually succumbed to the snare. Mitchell has actually co-authored a declaration with a variety of university AI departments requiring “a government-coordinated effort combining unions, universities and market professionals, the superannuation sector, and the Australian tech market to … construct our own AI designs.”

Obviously, the turkeys wish to begin designing for Christmas.

Quid professional quo

For the econocrats recommending Jim Chalmers in the Treasury, AI has actually become the long-hoped-for performance saviour that will stimulate a generational increase in financial results. Treasury’s just recently launched intergenerational report starts with a chapter dedicated to “the AI transformation”, with radiant descriptions of “a general-purpose innovation like steam power, electrical power and the web” that is “anticipated to have an extensive impact on our economy and society”.

AI exists as a technological remedy, with “the capacity to make vital services more available, cost effective and responsive to specific requirements”. While confessing some tasks may be lost, the report is extremely positive about the longer-term impacts, arguing that “as AI ends up being extensively embraced, it is anticipated to drive boosts in performance and earnings, supporting lower costs and greater customer need and work”.

This is the good sense underlying the breezy techno-optimism of the Productivity Commission and Treasury. It’s appropriately summed up by Stephen King and Julie Abramson, who, when they were PC commissioners, composed in The Australian in 2015 that “technological shifts– from computer systems to the web– are partially why the typical Australian today has living requirements 3 times greater than the typical Australian in the 1960s”.

In this worldview, pesky laws and guidelines like copyright are obstacles to the complete implementation of AI throughout the economy.

Artists and the cultural sector are battling tooth and nail versus such exemptions, which would permit AI business to train designs on imaginative works without payment. Anthropic and OpenAI are looking for the carve-outs for legal defense. Having actually enthusiastically pirated product to train their designs, they’re now dealing with the fallout: Anthropic, for instance, just recently handed over US$ 1.5 billion in a settlement to authors and book publishers.

In the wake of Amodei’s check out in late March, creatives came down on Canberra in a desperate lobbying effort to combat back. The federal government released suave guarantees that it would not alter copyright law. That hasn’t stopped extreme lobbying pressure from Anthropic and OpenAI as they work out behind the scenes.

The temptation of a quid professional quo– copyright carve-outs in return for information centres and AI training — will be strong in Canberra, enticed by the possibility of billions in financial investment and the “sovereign ability” of a home-grown frontier design. “Australia merely can not manage to miss this AI wave “, enthused Peter Hartcher in a column last monthdiscussing how Australia has the chance to protect access to frontier designs by permitting beneficial regulative treatment for Anthropic and OpenAI.

‘ Do an offer’

This concept, that the Albanese federal government need to “do an offer “with frontier AI companies on copyright

, has actually been taken up with gusto by Liberal Senator James Paterson in a series of media interviewsFor what, precisely, and for what in return, has actually not been defined.

” Anthropic and OpenAI have the most sophisticated designs worldwide, and what we must be looking for to make sure is that we have sovereign access to those designs,” Paterson informed News 24’s Andrew Clennell on September 27, declaring that this might be done by motivating the AI companies to invest more on training in Australia,” which will successfully ensure our access to these designs”.

Affordable concerns may be positioned about this putative offer, not least of them whether Paterson and his coworkers in parliament actually comprehend what a” frontier design” is. How would this offer be done? What does “gain access to” really imply?

Anthropic’s much-ballyhooed memorandum of comprehending with the Australian federal government, signed after Amodei’s victorious swing through Canberra this year, is an unclear nothingburger. It makes some airy pledges about” technical exchanges and partnerships with security and security organizations “and points out “Anthropic means to take actions towards guaranteeing advantages to regional market and scientists through access to its facilities”. The memorandum is not a binding offer, stating “this MOU is planned as a declaration of intent and is not planned to have legal impact. “

Thus much of the AI policy discourse, this MOU exposes more about the creative capture of Canberra’s elites than it does about the apparently historical chances or existential risks of emerging innovations. Little examination appears to have actually been used to the sales pitch. It does not take a great deal of creativity to comprehend why AI companies frantically going after earnings to validate their sky-high evaluations would be more than happy to provide Canberra bureaucrats a test drive on their newest items. Nor does it take the processing power of Claude to understand the windfall that a copyright carve-out would bestow.

The nationwide security argument is just a little more reputable. Possibly hyper-capable hacking representatives are operating on unique United States federal government servers someplace. Or perhaps the Australian Signals Directorate requires an unique agentic goalkeeper that it can’t construct itself. It’s not clear that “gain access toto the bleeding edge of design advancement is a requirement for Australia’s nationwide cybersecurity, even presuming Australian spy companies have the ability and functional ability to successfully utilize such designs.

Arguments that information centres physically situated in Australia can in some way be managed by the sovereign regional jurisdiction appear to show a touching faith in the probity of the companies running them.

On obtained money and time

At a more prosaic level, there need to be open concerns about whether any sovereign federal government can rely on the AI billionaires. Citizens are being asked to think that a Commonwealth that can’t handle to tax foreign-owned gas business will in some way change into a callous arbitrator for frontier modelling gain access to.

The Albanese federal government, agitated by Angus Taylor and Paterson, seems recommending it can strike beneficial hosting-for-access handle the exact same frontier AI business that has actually simply confessed to hacking into Australian federal government servers. As they utilized to state in the old days, it does not calculate.

It can’t be a surprise that normal people are fretted about a big and effective market executing a huge job of monitoring, automation, contamination and extraction. The magnifying opposition to widespread information centre building, both in the United States and significantly in Australia, is inextricably connected to growing stress and anxiety about the future threats and effects of technological modification. The market’s reaction has actually been to invest up on additional lobbying effortswhile the majority of the states and areas are still panhandling for every single brand-new information centre they can bring in.

The speeding up chance expenses reach their complete realisation in the hyperscaled information centre buildoutwhich has actually catalysed what can just be referred to as a totally blown speculative bubble. In a time when inequality is skyrocketing and billions do not have standard human security or access to tidy drinking watera large objective investing in non-human succour and assistance has actually been inaugurated that makes use of much of the world’s offered liquid capital.

According to current information assembled by Stijn Van Nieuwerburgh for the Brookings Institutionthe information centre build-out measures up to the financial investment put into a few of the biggest booms in historybigger than the canals and cotton mills of commercial revolution-era Lancashire, the American train boom of the 19th century, or the post-war American interstate highways funded by Eisenhower.

Much of this cash has actually been obtained. The medium-term incomes required to pay back these loans appear unpredictable, to state the least.

Maybe financiers truly do think that superintelligent device gods can change the white-collar labor force, vacuuming up swaths of the international economy’s everyday costs while doing so. The existing bubble likewise has all the trademarks of previous manias, in which a monetary sector in the grip of a speculative craze stacks into endeavors with just the flimsiest prospectuses.

The numbers can make the eyes glaze over, however some narrate of their own.

In August, Alphabet offered US$ 5.5 billion of business bonds in Australiawhich followed big bond programs in CanadaWhy is a business like Alphabet– up until a number of years back, a corporation swimming in complimentary capital– unexpectedly tapping smaller sized bond markets, obtaining billions of dollars in foreign currencies at vouchers above 6 %? The response is that Alphabet, like much of the rest of Silicon Valley, is investing stupefying quantities of cash into information centres: more than US$ 200 billion this fiscal year alone

Alphabet is abundant, with a long history of effective innovation develops. An entire class of lower-tier information companies has actually sprung up, with organization designs absolutely nothing more complex than guaranteeing to construct information centres with cash they have yet to protect. These are the so-called” neocloudsclimatic business developments that have actually obtained eye-watering total up to construct or rent information centre capability, which they then offer on to tech companies and other corporates wanting to get access to sophisticated processing systems.

Australian neocloud Sharon AI is an eyebrow-raising regional example. Helmed by James Manning, Sharon is noted on the NASDAQ and declares to manage 212 MW of information centre capability, the majority of it obviously rented from NextDC in AustraliaThe company has actually obtained more than $1 billion in convertible notes and just recently tattooed an oversubscribed $1.6 billion personal financing round. It seems utilizing this financial obligation to rent Nvidia AI chips.

In spite of its lofty aspirations, Sharon AI loses cash. According to its second-quarter monetary outcomesthe company scheduled simply US$ 1.9 million (with an “m”) in earnings in the 3 months to June 2026, for an operating loss of US$ 12 million. It does not appear to in fact own any information centres, noting simply US$ 26 million on its balance sheet for “home and devices”. Most of Sharon AI’s “contracted profits” is with a secret US$ 950 million client, whose name Sharon will not expose

Sharon is a photo of conservatism compared to its highest-profile regional rival, FirmusCurrently well-known for its breakneck advancement of Tasmanian information centres, Firmus is careening towards an IPO this month that some experts think might be among the biggest in the history of the Australian stock market.

Like Sharon and the American neoclouds, Firmus’ service method is a mix of vertical development projections and circular funding. Its essential provider is Nvidia, which is lending it cash to purchase its chips. Its crucial client is OpenAI. Regardless of this, Firmus has really couple of real chips, little presently existing information centre capability, and no openly divulged audited financials.

Oh, and among its creators, Oliver Curtis, is a founded guilty criminal who hung around in jail for conspiracy to dedicate expert trading. Unsurprisingly, opposition is growing in Tasmania, where Firmus prepares to draw up one-seventh of the state’s existing electrical energy supply

You do not need to be Charles Kindleberger or Hyman Minsky to see some looming threats. By Thursday, lurid rumours were distributing in business press that Firmus’ IPO remained in problemwith some hypothesizing that the company’s brokers would desert the preliminary offering entirely

‘Jobs and development’

Ask a common Australian what need to be constructed with a historical facilities program, and it’s a reasonable bet that “information centres” would be well down the list.

The majority of the country’s facilities requirements are social, not digital. Australia requires homes, aged care beds, budget friendly healthcare, schools for our kids, pits repaired and trains for our stretching external residential areas. We are just midway through a historical shift from a costly and contaminating coal-fired electrical power grid.

All of these requirements are taking on the abilities, products, and particularly the financing presently being put into server racks for Nvidia graphics processor systems. Possibly the most disturbing element of the AI bandwagon is the degree to which considerable financial investment choices about the future of the economy are being made by foreign billionaires, with little factor to consider of the requirements or desires of common Australians.

None of this appears to be fretting Australian political leaders, who are tipping over themselves to invite brand-new information centre financial investment, even as they declare to assure stringent brand-new safeguards on brand-new information centre advancement. Tasmanian Premier Jeremy Rockliff has actually been an passionate spruiker for Firmuswhich is sponsoring the brand-new Tasmanian Devils AFL football franchise, while in New South Wales, Premier Chris Minns has actually presumed regarding declare that the state would remain in economic downturn if it weren’t for information centre costs.

The endearment of political leaders for “tasks and development”is one factor for the large gulf that has actually opened on AI in between elite and popular viewpoint. While the technocrats imagine endless chances at the modelling frontier, normal folk can’t assist keeping in mind of the rhetoric of task damage and human termination spouted by the Altmans, Amodeis and their mid-level San Francisco lackeys.

Business laying off personnel are now regularly declaring AI is accountable, consisting of Atlassian in Australia, while the media has actually enthusiastically covered both current “rogue representative” security attacks and the wilder declares of the reliable altruist doomerists. Mild-mannered suburbanites are banding together to eliminate the information centres emerging in their yards, while none besides the Reserve Bank is blaming the information centre costs for treking rates of interest.

Sociological, mental and humanitarian theories about innovation assist us comprehend the weird times we are enduring. Effective brand-new innovations are exceptionally political developmentseven as their wider effects are socially and financially disordering. Australia’s political class has actually regularly had a hard time to come to terms with emerging innovations and has actually usually picked to react to the obstacle by outsourcing its policy concepts to tech giants and consultancies, and AI is just the current case.

What’s at the heart of Albanese’s task

By mid-2026, the dispute about AI had actually reached a phase where Anthony Albanese plainly felt the requirement to make a public declaration, if just to kick the can down the roadway till some real guidelines might be prepared.

In July, the prime minister provided a speech staking out his position. Entitled “AI in Australia’s interestsAlbanese tried to stabilize the chances and hazards of AI while worrying a frame of nationwide interest. “Our power, our firm, our option depends on welcoming modification and forming it,” he argued. “Not simply embracing or accommodating AI: developing it, making it, developing the ability right here.”

In the speech, Albanese devoted to securing copyright and putting in location “the greatest possible security for Australian artists and Australian media”. Weeks later on, ACT independent Senator David Pocock tabled private assessment files proposing an “opt-out” copyright exemption for design training, which lots of in the innovative markets think was covertly worked out with OpenAI and Anthropic.

An “opt-out” provision would enable innovation companies to immediately utilize copyrighted product for training, unless artists and copyright holders clearly object, producing a loophole that would enable AI companies to lawfully utilize scraped material for their training. This has actually started a round of anguished op-eds from artists and authors like Richard Flanaganhowever the hand-wringing might not suffice to stop the AI lobbying juggernaut.

In her report for The Saturday Paper todayKaren Barlow estimates Labor backbencher Ed Husic (a previous Industry minister himself) explaining a “shock and wonder lobbying method by United States tech, developed to overload federal government decision-making”. Kate Pounder, a previous employer of the Tech Council of Australia and among Andrew Charlton’s previous coworkers in the lobbying company AlphaBeta, has actually simply been designated by the Malinauskas federal government as one of the commissioners for its state-based royal commission on AIPounder has actually likewise worked for OpenAI. It’s not looking helpful for the artists.

The within video games shed even more light on why Albanese and his federal government are still speaking with Sam Altman, even after an OpenAI representative acquired unauthorised access to a Medicare website previously this year. OpenAI has actually been less than upcoming on the occurrence: Altman met Deputy Prime Minister Richard Marles in early September, where he in some way forgot to discuss the attackSenior executive Jason Kwon, who’s been sent for a round of apologies, declares Altman didn’t understand about the hack when they satisfied. Regardless, the event does not appear to have essentially modified Labor’s interest for frontier design gain access to.

The federal government’s indecision about whether to manage AI mirrors a wider confusion at the heart of Albanese’s task. Given that 2022, Labor has actually tried to execute a brand-new design of market policy, funding a big mutual fund for a “Future Made in Australiaand bailing out this or that smelter or refinery. There appears to be little in the method of overarching financial method, and many of the basic settings are the exact same neoliberalism of the previous 40 years. In this context, the federal government appears to be dealing with AI like a digital variation of the mining boom– a contrast drawn by Charlton in a speech to ANU in August. “We require information centres, however just the great ones,” he declared last month.

AI policy demonstrates how Labor’s wider social task under Albanese has actually gradually burrowed. The celebration appears too separated from its deteriorating electoral base to react to the clear desire from numerous working Australians for a time out in the AI hype-doom loop.

A federal government in the sway of tech lobbyists, laced in the United States nationwide security device, moving quickly to execute brand-new algorithms to identify aged care gain access to and making severe cuts to special needs care is not most likely to be one that will control AI in the very best interests of common Australians.