The EY survey released this month put a number on how fast the world is moving: 16% of people globally have already used AI that acts on their behalf (buying products, managing finances, booking appointments) without human intervention. In Pioneer Markets that figure hits nearly 25%. AI is not inching toward autonomous decision-making. It is already there, already running.

Australia is running consultations.

What the EY data actually says

The report, “As AI moves from advice to authority”, is not speculation. It is a survey of current behaviour across dozens of markets. Ten percent of respondents let AI purchase products for them. Eleven percent let it manage banking tasks. Twenty-one percent would let it book medical appointments automatically.

The question it asks is who defines the limits of autonomous AI. That question determines which economies capture the value and which ones miss the moment entirely while writing frameworks about it.

The Australian record

In September 2024, the Australian government proposed mandatory guardrails for high-risk AI. A year later, those guardrails had not progressed. In August 2025, the Productivity Commission recommended pausing the mandatory guardrails work entirely, calling new AI-specific laws a “last resort.” Commissioner Stephen King’s exact framing: “Adding economy-wide regulations that specifically target AI could see Australia fall behind the curve.”

In December 2025, the government released its National AI Plan. Multiple legal analyses called it big ambitions, light on details. The regulatory section replaced the proposed guardrails with a philosophy of regulating as little as possible. The government committed AUD 29.9 million to launch an AI Safety Institute.

The UAE committed to building a university whose entire research focus is artificial intelligence.

What the UAE actually did

MBZUAI — the Mohamed bin Zayed University of Artificial Intelligence — sits ranked among the top 20 universities globally for AI research publications. It has 700 students and alumni from 49 nations, over 100 faculty drawn from the US, China, and Germany, and offers fully-funded scholarships. A satellite research facility opened in Silicon Valley in May 2025. Roughly 80% of graduates stay in the UAE.

That last number is the point. MBZUAI is not a PR asset. It is a talent retention machine.

The Dubai International Financial Centre has announced it will become the world’s first AI-native financial centre, embedding AI into its legal frameworks, business environment, and physical infrastructure. DIFC now hosts over 1,600 fintech, AI, and innovation firms. Abu Dhabi has over 400 AI companies, up 41% year-on-year. The UAE’s AI market is projected to reach USD 4.25 billion by 2033, with AI targeting 20% of non-oil GDP by 2031.

And businesses in UAE free zones can be licensed in as little as 60 minutes, fully online, no office visit required.

The wrong comparison to make

The standard defence of Australia’s approach (and most Western democracies’) is that the UAE is an authoritarian state and democratic governance takes longer. True. Irrelevant.

The relevant comparison is outcomes for founders. A founder building an AI product in 2026 is choosing where to incorporate, where to hire, and where to raise. They are looking at corporate tax rates (zero in UAE free zones versus 30% in Australia), setup timelines (hours vs. months), regulatory clarity (the UAE tells you what you can do; Australia is still deciding), and access to capital (UAE companies raised USD 1.2 billion in Q3 2025 alone).

“But democracy is slower” is not a competitive moat. It is a concession.

The safety delusion

Countries slow on AI adoption tend to justify it as responsible. They are protecting citizens. They are doing due diligence. They are not like those reckless places.

What they are actually doing is exporting value. When founders and engineers go where capital is freer and regulation is lighter, they take the compound interest of their work with them: the know-how, the networks, the tax base, the next company they build. The EY report makes the trust dynamic plain: seven in ten people want human oversight of autonomous AI, yet 16% are already using it to manage their finances. Behaviour precedes confidence. Markets follow behaviour.

The economies that define the defaults for agentic AI (the ones whose companies build the products that become the infrastructure) will set the norms for everyone else. Not by decree. By the weight of what already works and who built it.

The window for “we’ll get to it” is closing.

What Australia should actually do

Not more consultation. Not another framework. Not a safety institute that launches 18 months after the Productivity Commission told you to stop writing regulations.

Australia has real advantages: strong universities, English-language access to global talent, time-zone overlap with fast-moving Asian markets, and a legal system capable of handling complex commercial arrangements. None of that matters if founders keep leaving.

What moves the needle: cut business setup time to days. Create genuine AI-focused special economic zones with tax incentives that compete with DIFC and ADGM. Fund an AI research institution with the urgency the UAE funded MBZUAI. Stop treating AI regulation as a risk management exercise and start treating it as economic development.

The EY report’s core question is who defines the limits. The answer is practical: the people who showed up. Right now, Australia is still deciding whether to come to the meeting.

If you are a founder working out where to build and how to compete in an agentic world, talk to us.

Sources