Why explicit deepfake content feels impossible to stop
By Raidah Alam
Imagine discovering a video online that appears to show you in an intimate situation you never consented to. And one you were never actually a part of.
If you have ever come across a deepfake video, you will know just how perturbing yet convincing they can be. But this is no longer a distant risk tucked away in a strange corner of the internet. Advances in artificial intelligence have made it possible for almost anyone with an internet connection to generate highly realistic ‘deepfake’ content, often using nothing more than a few publicly available images. You no longer need to be technologically skilled to create this content, nor do you need much more than a simple online presence to potentially fall victim to it.
And while there is now widespread awareness of these risks, this content continues to proliferate out of control. Data indicates that explicit deepfake material online has surged by as much as 550% each year since 2019. It is reported that its growth rate surpasses almost all other cyber threats.
So as synthetic media continues to spread at speed, the law faces an urgent question: how do we protect individuals in a world where seeing is no longer believing?
Intellectual property law has been considered as a possible avenue as it is designed to regulate ownership and control over content. But whether it can meaningfully protect individuals from this form of harm remains uncertain.
Get Kylie Pappalardo’s stories in your inbox
Join Medium for free to get updates from this writer.
In the meantime, in Australia, this is not a wholly unregulated space. Federal law[1] criminalises the distribution of non-consensual explicit material of an adult without their consent, including material that has been created or manipulated using artificial intelligence. Despite these laws, the issue still feels far from resolved. Why has there been only one federal prosecution of its kind, when the problem remains so widespread? This likely reflects how difficult these cases are to pursue in practice. There are jurisdictional limits, challenges with evidence, and a high bar for criminal liability.
The law touches part of the problem, but not the whole experience of losing control online. The scale and global structure of the internet itself makes enforcement incredibly challenging. Evidence disappears while content is multiplied in seconds, often across multiple platforms or countries at once. Even if there is a legal action to pursue, it is dependent on complex investigations requiring digital forensics, cooperation between countries, and access to platform data. These resources are very difficult to coordinate.
We cannot rely on the platforms themselves either. They frequently avoid involvement or responsibility for user-shared content by hiding behind their position as ‘intermediaries’. This is unlike the urgency around copyright infringement, where platforms tend to act quickly to remove content and absolve themselves of liability. As a result, a cruel burden is placed on victims. They are expected to navigate a patchwork of international platforms and keep reporting content while it continues to multiply elsewhere, anew.
There is also an issue in how the law itself is written. Legally, it is irrelevant whether the explicit material is unaltered or altered using artificial intelligence. This means a real image and a wholly AI-generated deepfake are treated in the same way. But they are not the same harm; one is a violation of privacy while the other is the creation of an entirely false reality. This crucial difference is flattened by the law.
So yes, Australia has made explicit deepfake material illegal. But for many people, the problem is not in its legality. It is in how quickly it spreads, how difficult it is to remove, and how little control over one’s own image exists once it appears distorted online.
The legal system is playing a cat-and-mouse game with deepfake technology. If it does not find another approach, it will be permanently outpaced by this burgeoning target.
What do we really need? Because it is clear that this reactive model is not doing enough for deterrence. Legislation needs to proactively focus on consent, strict liability for perpetrators, and immediate removal obligations for platforms. Intellectual property law may be a better place for this protection. Copyright is notoriously skilled at making big tech companies listen to its demands.
If you hear a 5-second snippet of a copyrighted song in the background of a 30-minute YouTube vlog, you will see the clip muted or the content taken down very, very quickly — almost instantly when rights are triggered. The question is: why does similar responsiveness not exist for deepfake abuse? Because at its core, this is about losing control of your own image in a system that can recreate you, without you.
[1] Criminal Code Amendment (Deepfake Sexual Material) Act 2024 (Cth); Criminal Code Act 1995 (Cth) s 474.17A.
