‘Supervised’ self-driving cars are here – and Australia’s laws aren’t ready. Here are 3 ways to fix them

In September, US electric car maker Tesla rolled out a semi-autonomous driving feature it describes as “the future of transport” in Australia.

As its name suggests, the Full Self-Driving (Supervised) system blurs the line between human and machine control. Our current licensing and road-safety frameworks were not designed to handle this situation.

A federal government strategy for high-tech road transport released last week has little to say about how this new semi-autonomous technology should be managed.

As experts in cities and transport, and how people use them, we have some concrete policy ideas for how to manage this innovation safely: changes to licensing rules, safety testing, and accountability and transparency.

What is ‘supervised self-driving’?

Tesla’s new system is the most advanced form of semi-autonomous driving yet available to Australian consumers.

The car can follow routes from start to finish, handle intersections, change lanes, and respond to traffic lights. These tasks go well beyond traditional “autopilot” or adaptive cruise control.

Tesla says that “under your active supervision”, the system “will drive you almost anywhere”. Despite “full self-driving” in the product name, the active supervision part is crucial.

Early testing in Australia by reviewers from The Drive and CarExpert shows a technically impressive system that is nevertheless far from flawless:

• the car recognises signs but doesn’t always read or reliably interpret them
• lane changes can be slow, sometimes confusing nearby drivers
• navigation can falter when satellite coverage is weak
• the system occasionally selects the wrong lane or misjudges right of way
• it may park awkwardly, at times straddling two bays
• wearing dark sunglasses can confuse the camera that monitors the driver, and disable the system.

The overall impression is of a vehicle that can handle itself most of the time, but which needs an actual driver ready to take over at a moment’s notice. And to be fair to Tesla, the company does say the system “requires an attentive driver to maintain proper control of the vehicle who must be ready to take immediate action at all times”.

Is Australian regulation keeping up?

The federal government’s new National Road Transport Technology Strategy outlines how Australia plans to prepare for connected and automated transport.

The strategy commits to developing a “consistent national regulatory framework” for automated vehicles. However, it notes that implementation will rely on state and territory transport strategies, which set out more detailed plans for adopting and deploying new technologies.

In short, the strategy recognises automation but says little about how it will be governed.

What policies could work?

Supervised automation occupies a grey zone between human and machine control. Australia’s current policy framework has yet to define how that space should be managed.

There is no single solution, but several areas stand out where modest, evidence-based steps could help the supervised driving technology mature safely.

1. Driver training and licensing reform

Supervised automation changes what it means to drive. The driver’s role shifts from active control to continuous monitoring. Research shows people are not naturally good at this.

When drivers are not required to maintain continuous control, their awareness of their situation can erode quickly. This may lead to slower or poorer reactions when they need to take control.

Modest updates to existing licensing tests could help to address this. State authorities could introduce additional knowledge questions or a short modules covering these systems and their limitations.

These would ensure every driver using such systems has at least a basic understanding of how they work, and where human accountability begins and ends.

2. Pre-deployment certification and transparency

Before semi-autonomous systems reach the public, they should demonstrate they can perform safely under local road, climate and traffic conditions.

At present, Australia has no requirement for such validation. Once a vehicle meets the Australian Design Rules, manufacturers can activate or modify automation features via software updates. They do not have to submit supporting safety data or seek regulatory approval.

Manufacturers could be required to provide validated performance data before releasing updates. This might include things such as how often and how well humans take over from the system, how often sensors fail and what happens afterwards, and measures of how effectively the system avoids crashes.

Trials of “beta” or experimental software could still be possible, but only within regulator-approved test zones under controlled monitoring.

A complementary step would be to embed minimum performance benchmarks for these systems within the Australian Design Rules or ANCAP safety-assessment protocols. These might cover the timing of alerts, the reliability of the system, and transparency of about the system’s performance.

3. Accountability and data integrity

At present, Australia has no requirement for semi-autonomous car manufacturers to share data on crashes, near-misses, or times when the system turns off or hands over to the human driver.

When incidents occur, the evidence often stays solely with the company. This limits public scrutiny and independent investigation.

Here, Australia can learn from regulations in other jurisdictions.

For example, California publishes annual “disengagement reports” from all developers testing automated systems. The European Union requires manufacturers to retain and share data from critical safety systems for crash investigation and oversight.

Several legal and regulatory questions also remain unresolved, regarding who is responsible for crashes, what drivers’ obligations are, and how insurance should work.

What now?

Australia faces the task of making rules for the in-between space of “supervised self-driving”.

The technology itself is no longer experimental. It is active on our roads. What’s experimental is how we govern it.

Building clear, modest guardrails now will help ensure that automated mobility develops safely. We shouldn’t wait for major incidents to reveal the gaps and create a drive for hurried regulation.