Real AI policy is over there
In the mess of rules that will block AI abundance
Most governments think that good AI policy is imagining disasters and designing ways to slow down progress: traffic light systems, alignment with “core socialist values”, mandatory killswitches. But real AI policy isn’t here.
Real AI policy is over there — reforming the mess of outdated regulations that will stop us from putting AI to practical use. For decades, AI-driven possibilities will collide with highly regulated jurisdictions.
Some worry about where AI models are trained and run. But the jurisdictions that thrive will be those willing to undertake deep economic reform to unlock AI’s potential. Only then will we get richer and freer.
I hope Australia is one of those places.
Nice things we won’t be allowed to have
What happens when AI discovers thousands of bespoke medicines and medical devices?
The cost of a successful submission to the Therapeutic Goods Administration (TGA) is between $500,000 and $1 million, with potential delays stretching for years.
What is the point of medical breakthroughs if regulations prevent their use? We can’t just discover new medicines, we need to take them.
Today’s free open source models are capable of sophisticated personalised financial advice.
The Australian Securities and Investments Commission (ASIC) requires financial advisors to be licensed to provide personal advice. Yet applying human-centric compliance rules to foundation models makes little sense.
Will AI models need licenses? Most AI models have probably already read more financial regulations than human advisors.
Legal help remains unaffordable for many Australians — not because AI can’t provide it, but because it is a crime to provide unqualified legal advice. AI platforms can’t be qualified lawyers, so any company offering AI-driven legal services risks breaking the law unless a licensed lawyer supervises every interaction.
Students pay unsustainable fees for an education model that hasn’t fundamentally changed in centuries, while regulatory barriers block AI-driven alternatives.
AI models are a remarkable teaching tool at very high levels of education. Yet the Tertiary Education Quality and Standards Agency’s (TEQSA) rules require “academic teaching staff” with credentials. How could an AI-taught course meet Australia’s Higher Education Standards? It won’t, even if it’s better.
What about out autonomous vehicle future? Australia’s delivery drones still require a licensed and certified drone operator, enforced by the Civil Aviation Safety Authority (CASA). Doesn’t sound very autonomous to me.
These are all AI policies.
Deep reform
A forward looking AI policy agenda means deep economic reform. Not just tweaks, but hard, controversial reforms.
Australians will not benefit from AI if they fixate on hypothetical risks. We must overhaul and rebuild the regulatory frameworks. We need regulatory sandboxes, right-to-try policies, mutual recognition agreements, and exemptions from occupational licensing. And that’s just the start.
The real work of AI policy does not look like AI policy at all — it lies hidden, buried in the countless regulations obstructing AI abundance.