Australia is preparing what could become one of the toughest AI regulatory regimes in the world. Prime Minister Anthony Albanese announced that mandatory national standards are coming, targeting two things AI companies have mostly gotten to use for free: electricity and creative work.
For years, Australia treated AI the way it treated most new technology, applying existing laws and letting voluntary guidelines fill the gaps. The Privacy Act, consumer protection law, and anti-discrimination statutes covered AI decisions much as they covered human ones.
Australia also introduced a Voluntary AI Safety Standard and a government policy on the use of AI, most recently updated in December 2025. However, neither is legally binding.
That is expected to change in early 2027 with new legislation that will create an Office of AI within the Department of the Prime Minister and Cabinet. The office will develop legally enforceable AI standards instead of voluntary guidelines.
The most concrete proposal targets data centers, and it inverts their usual role. Instead of simply drawing power from the grid, new facilities will be required to become net energy producers, feeding in at least as much as they consume.
They will also have to fund their own power connections so other electricity users are not left covering the cost, meet mandatory water efficiency standards, and help pay for whatever extra water infrastructure their operations demand.
READ: Kenya’s AI Policy, As the Government Sees It
Communities get a say in where these facilities will be built. Given how much water and power AI infrastructure is projected to consume in coming years, especially around Sydney, the government wants these obligations settled before scarcity becomes the story, not after.
The second pillar concerns copyright. Australia is rejecting the broad text-and-data-mining exemptions that let AI firms in other markets scrape copyrighted material with little friction.
Under the proposed law, writers, musicians, artists, and journalists keep ownership and control of their work, including the right to set its price.
Albanese put it bluntly:
We say that it’s important that creators get paid for their work. Writers, artists, and musicians should remain in control of their work. They must give their consent and be fairly compensated if their work is used. Anything less than that amounts to theft.”
What remains unresolved is enforcement. Licensing could run through individual deals or collective schemes, but questions linger about offshore training, models already built on Australian material, and who polices any of it.
The EU’s AI Act is risk-based and still permits text-and-data mining under certain conditions. The US relies on executive orders and voluntary company pledges instead of one federal AI law, letting AI firms use copyrighted material for training without asking first.
Australia takes a different approach by regulating the energy and creative materials used to develop AI, rather than focusing only on AI products once they reach consumers.
Kenya’s picture looks less advanced by comparison. The Artificial Intelligence Bill, 2026, is still before the Senate, and it follows the EU’s risk-based tiers rather than Australia’s infrastructure-first model.
It would create an Office of the AI Commissioner to classify high-risk systems, but it says nothing about data centers, energy, or water, largely because Kenya’s National AI Strategy 2025-2030 treats infrastructure as something to build up rather than restrict.
The gap around copyright is even more significant.
READ: Judiciary Moves to Regulate AI Use in Courtrooms After Fake AI Citations Surface in Court
Kenya’s Copyright and Related Rights Bill, 2026, modernizes the old 2001 Act and strengthens the Copyright Board’s enforcement powers, but it does not touch AI training at all.
Kenyan songs, photos, journalism, and code are already being scraped into datasets with no payment or consent required, and the pending bill leaves that gap open.
Australia can regulate AI at the source because it controls valuable resources like land, energy, and water that data center operators need no matter the regulations. Kenya, on the other hand, is still trying to attract data center investment as part of its digital growth.
That gives it less room to impose strict requirements without risking investors choosing other countries. This difference in bargaining power likely explains the different approaches more than any difference in political commitment.
AI is expected to contribute about AU$116 billion (~ KES 10.5 trillion) to Australia’s GDP over the next decade, making a strong case for tougher regulation.
Kenya’s AI sector is also growing rapidly, but until its AI Bill is passed by Parliament and its copyright laws address the use of training data, the country is likely to continue regulating the impact of AI rather than how AI models are built.
