Protecting jobs by killing growth

If you were hoping for the Australian economy to continue benefiting from the AI revolution, this should be terrifying reading:

"Prime Minister Anthony Albanese is actively involved in developing the new suite of AI policies, according to government sources not authorised to speak publicly. His intervention comes as unions push state and federal Labor to do more on workers' rights, and Labor MPs face growing local pushback against rapid data centre expansion.
...
The government has made clear that it expects new data centre projects to add electricity supply to the grid to cover all or part of their power usage if they want fast regulatory approvals.

But Labor is now also considering imposing stricter requirements on data centre operators, such as forcing them to provide financial benefits to local communities where they set up operations, rather than mere non-binding expectations."

Labor's draft policy platform also includes what could be a union veto over use of AI in the workplace, with employers required to "meaningfully engage" with workers "about the introduction and use of artificial intelligence and automation in the workplace".

I get the intent; AI is a big, new scary technology and fearmongering about economic disruption and job losses is rife. But all of the 'solutions' leaked above are economically misguided and, if enacted as policy, will do far more harm than good.

Let's start with the involvement of the Prime Minister himself. Albo has proven himself to be something of a Luddite in terms of technology, with one of his signature legislative achievements being the Social Media Minimum Age Act, an ill-conceived piece of regulation that doesn't even work because it misunderstood the technology. I'm not sure what Albo's increased involvement in AI regulation is exactly, but given his tendencies towards technology and the fact he's a member of the CFMEU-aligned Labor Left faction, presumably it means a heavier "hands-on" approach, specifically with regard to the labour market.

That might look well-meaning. But unions don't care about Australian workers; they care about their members. Specifically, they care about protecting their members' existing jobs. In the face of a technological shock, the interests of incumbents and the broader workforce can diverge. Regulations designed to give union workers more power to protect jobs would do a disservice to workers, especially those who might use AI to disrupt entrenched, union-backed patterns of exchange.

If Australia wants its labour force to be able to adapt to AI, it needs to move the regulatory needle away from unions and job protection and towards something more closely resembling the Danish 'flexicurity' model where the worker, not the job, is what's protected.

As for requiring "new data centre projects to add electricity supply to the grid to cover all or part of their power usage", that's the heavy hand of government misunderstanding how markets work. It's true that Australia is experiencing something of a data centre construction boom (although it pales in comparison to the US), and that's having a real impact on people and infrastructure. But markets, when allowed to function through price signals and voluntary exchange, are remarkably good at solving these problems.

For example, network costs (the poles and wires) in the energy grid can be up to half of an Australian household's energy bill. Adding data centres into the mix essentially splits that cost over more users and more hours (data centres provide consistent, 24/7 baseload demand), so while total usage goes up, retail prices can actually come down. Requiring that they cover their own power usage is a highly inefficient allocation of capital that ignores comparative advantage and relies on clunky government mandates rather than wholesale price signals, which are far better at incentivising the most efficient new generation on the grid.

As for "financial benefits to local communities", that's dangerously vague. What counts as a benefit? How is it measured? Such vagueness is a recipe for giving excessive discretion to regulators and inviting local rent-seeking and exactions, creating uncertainty for investors, which is essentially a tax on innovation. Data centres already pay for their grid connections, the energy they use, and taxes; in Virginia, data centre revenue has grown so much that they've been able to reduce real estate tax rates for households every year for the past decade, while also supporting schools and government services.

Government mandates pre-empt such solutions and replace them with less efficient, politicised outcomes. They risk tying the sector up in regulatory knots, imposing costs and uncertainties that deter the very investment needed to power it.

Australia is a small, open economy. Realistically, it's not going to build the next OpenAI or Anthropic. But it doesn't need to. The policy goal should be to make Australia the world's fastest and most sophisticated adopter of AI, driving productivity gains across existing and new industries. If Albo wants Australia to take part in that AI revolution, he should be looking to slash the many layers of regulatory approvals and bottlenecks that have caused timeframes for major projects to exceed 690 days in NSW. AI and data centres aren't commodities sitting in the ground; the market for AI is global. That means to take part and capture its benefits, you should be making it easier to build and adopt, not adding new regulatory hurdles that effectively trade away long-term growth for short-term political appeasement.