David Shoebridge (NSW, Australian Greens) | Hansard source

Thanks, Acting Deputy President; I appreciate your efforts. That's pretty remarkable, isn't it? One Nation comes in here and defends the right to create deepfakes, and then tries to draw a parallel between the appalling harm that overwhelmingly women are facing with these appalling sexualised deepfakes that are able to be produced through AI and political satire or picturing somebody as being a member of a band when they're not a band. To come in here and minimise the reality of deepfakes—particularly how they're used in the manosphere and the misogynistic way in which they've been used across the internet—and belittle it, like Senator Hanson did, is an extraordinary act.

Of course One Nation is going to come in here and defend deepfakes. They want to keep running their big offshore deepfake factory in Vietnam. They want to be able to keep getting the dark money from wherever the hell it's coming from. Whether it's coming from Russia, a tech bro in the US, Gina Rinehart or wherever they get their funds from, they want to be able to keep the funds coming in to produce the deepfakes in order to lie and spread their racism and misogyny. That's what they want. They want their deepfake factories to keep coming in and destroying any sense of collective response. That's why One Nation comes in here, and that's why Senator Hanson opposes this bill, the Online Safety and Other Legislation Amendment (My Face, My Rights) Bill 2025.

What I find remarkable is we don't have the government trying to find a pathway to get this bill through or to say deepfakes, without consent, should be illegal in this country unless they meet criteria such as those that already exist in the Copyright Act. Pick up those provisions in the Copyright Act that exclude satire—exclude those existing exemptions in the Copyright Act. We know that would work. We could just pick them up, pop them in this bill and have very clear exemptions. But, instead, we're going to get to this point where, literally, the Labor, Liberal and National parties all join with One Nation to vote down a bill to stop deepfakes.

We're watching One Nation with their right-wing politics using this technology to literally tear our society apart, to spread deep lies about politicians and about political positions, to spread deeply destructive messages online: racism, hate and lies. We have a chance here with this bill, to take it through to committee, to work through some amendments and to say, 'Actually, the hate factories of One Nation, in the right of politics, should be illegal when they're abusing people's images and producing realistic deepfakes.' But, instead, Labor's just shutting it down; I've got to vote it down today. There's so little leadership, again, from the Labor government on a core issue.

Now we have One Nation making the false argument that this bill would stop their 'please explain' cartoons. Well, I'd recommend One Nation read the legislation before they come in here and speak about it, and I think that's generally good advice for politicians before they speak about legislation. What is the definition of 'deepfake' in this bill? It is material that could be a still image, a speech, music or sound. And then it says, very clearly, 'the material depicts, in a realistic way, an individual's face or voice or an attribute or aspect of an individual's face or voice'—in a realistic way. This isn't picking up political cartoons; it's not picking up satire like that. The Greens wouldn't support legislation to do that. I'll tell you again: when you have politicians come in here and lie about the legislation, deliberately misleading people about the effects of this legislation in the course of trying to retain the right to spread deepfakes against people, I think it exposes exactly what that political project is for.

What does this bill do? Well, let's take a step back. It's not in dispute now that AI can fabricate a person's face, their voice and likeness convincingly enough to deceive people, to humiliate people, to exploit and profit. The cost of that falls on the individuals who are often not politicians. Often, they're just people trying to live their lives without being humiliated, insulted or misrepresented. The ability of AI to do that is continuing to improve.

This bill goes to something essential about being human. Our face, our voice, our presence, the way we walk—it goes to those singular and essential natures that each of us know is us as an individual, as a human. Protecting that essential humanity, not just our data or our reputation but what makes us us, is what's at stake here. It's a simple principle, and I'm surprised we can't get agreement on it across the chamber: that your face, your voice, the way you walk and your likeness are yours, and they can't be appropriated by somebody and used without your consent. And they definitely can't be appropriated by somebody and used without your consent to drive the hate, the misogyny and the profits of overseas big tech. And, unless we deal with this, that's what's going to keep happening. So I welcome Senator Pocock's bill and thank him for bringing the legislation in. It's a meaningful step to address a real and growing harm.

The bill would strengthen the Online Safety Act with a dedicated complaint system for deepfake material and would give the eSafety Commissioner powers to issue take-down orders on hosts and on users to take down deepfake content once a complaint is received. It also allows for consent to happen, and there actually are instances where what would otherwise be a deepfake may have a really powerful role if it operates with consent. For example, some people are using AI to create what would otherwise be a deepfake to do real-time translations in multiple different languages. If that's done with consent and it's getting out there and it's helping communicate issues across a diverse community, there are arguments that that can have a positive outcome. But it needs to be with consent, and I think it should be labelled as deepfake content so people can tell that that's what the material is.

The bill also creates penalties for posting non-consensual deepfake content on social media and lets the commissioner direct that fast removal and have the platforms held to account. It's a really important provision. It creates a new private right, a tort, where individuals can seek injunctions, damages and to have the harmful material removed. In doing that, they don't have to go to the expense of proving financial loss. It's assumed that a non-consensual deepfake is harmful and there's a right to have it removed.

The bill references human rights obligations under the ICCPR and the UN Convention on the Rights of the Child and preserves legitimate use by journalists, law enforcement, ASIO and scientific research. I want to be clear: if this bill gets to committee, the Greens would also like to see incorporated in this bill those exemptions that exist in the Copyright Act that allow for satire and similar reasonable use provisions.

It's critically important to give people impacted by deepfakes the power to have them removed and to seek a remedy. It's also critically important for our law to seek to work upstream. I don't often quote the Pope, but I reckon Pope Leo was useful in the first encyclical this year, where he talked about how technology takes on the characteristics of those who devise, finance and profit from it, and its worth must be measured by who it harms and who it serves. I think that's a really important contribution to how we look at technology: who does it harm and who does it serve?

The deepfake platforms, particularly the nudify apps and the platforms monetising them, aren't neutral infrastructure, and we shouldn't regulate them as though they're neutral. We should recognise the harm and we should regulate to protect people from the harm. We should recognise who it serves and we should regulate to hold them to account. The tools and platforms that generate this material currently face no real accountability, and they keep profiting. So we need that safety by design and those duty of care obligations to work upstream on generative AI providers to really protect communities. Theft or abuse shouldn't only be unlawful if you're caught or you target someone who is a lawyer.

It's the Greens' position that, until laws comprehensively deliver that safety, platforms that are shown to present an inherent risk, particularly of deepfakes, should be banned. We could start with Grok, and we could work on from there—we bloody well could. A take-down and sue model is at the heart of this legislation. Whilst it does create a right and we think it's a powerful way to create a right, I think we should be having a look at alternative ways in addition to that to ensure that deepfakes can be addressed. Giving a more proactive right and a more proactive regulatory power to eSafety is one of the ways of doing that. Of course, whilst I recognise that torts are an important way of holding people to account, it often requires money and lawyers, and we want to make sure that everybody who's impacted by deepfakes has a realistic right to get a remedy.

The take-down model also captures the public viral nature of some of this material. But we need to make sure it also reaches encrypted messaging, peer-to-peer transfer, some of those closed forums and offshore sites. I think that's a difficult task, but I think public sharing is only one dimension in which deepfakes can actually cause harm. We should be looking at that, and the government should be working with the crossbench in order to come up with provisions that don't only look at public sharing but look to some of that other non-public or less public sharing where we know that there's coercive material and we know that this is causing harm. With sextortion, in particular, the threat to share is the weapon and the image may never be posted publicly. That needs to be picked up in this legislation.

I'll say it again. I appreciate Senator Pocock for bringing this to the chamber. In the lead-up to the last federal election, the Greens were very clear about saying we need to get ahead of this and we need to prevent deepfakes from poisoning our politics. I saw Senator Pocock at that same time raising the issue in real time before the last federal election. Of course, the government didn't listen.

I want to be clear: there are jurisdictions that are forging ahead and getting this far closer to right than Australia does. You could look at what South Korea has been doing in regulating its online platforms and preventing deepfakes being shared. Look at some of the creative regulation that's happening in Taiwan about ensuring the online space is meeting basic duties and having some nimble regulation. There are places in the world we can look at to get working real-time examples about how we can address deepfakes while keeping the right to produce satire and comedy material and to hold politicians to account. We can keep the right to do all of that but come up with laws that say: 'You can't use deepfakes for misogyny, racism and hate. You can't use deepfakes without people's consent. You can't see politicians and political parties creating these hate factories and attaching them to deepfakes.'

I say again that I urge the government to vote to put this into committee and then to work on amendments, if they're needed, to put in the copyright provisions to take this forward, because simply voting this down today—walking across and voting with One Nation to vote this down today—will send exactly the wrong message. It will send the message that the deepfakes, the hate, the misogyny and the racism are not going to be addressed under the Albanese Labor government. Let's not get there. Let's support the second reading of this bill. I again thank the mover of this legislation for bringing it to the chamber.