Imagine scrolling through social media during an election campaign and coming across a video of a political candidate saying something shocking, hateful, or completely out of character. Your heart races; you immediately share it, add a comment of disgust, and keep scrolling. Later, a fact-checker reveals that the video was a deepfake—a synthetic creation made with artificial intelligence. The candidate never said those words. The event never happened. This is no longer science fiction. In recent years, the technology for generating hyper-realistic audio, images, and video has become widely accessible, and anyone with a laptop, some software, and a connection to the internet can produce deceptive content that looks staggeringly real. In the electoral context, this is terrifying. A candidate can be made to appear at a rally they never attended, to endorse a policy they have spent their whole career opposing, or to confess to a crime they never committed. These fabrications can spread far more quickly than any correction, and the consequences for democracy are profound. Deepfakes mislead voters, distort political preferences, and cast a fog of uncertainty over everything that voters see and hear. When people cannot tell whether a video is genuine, trust in the electoral process begins to crumble. And once trust is destroyed, it is extremely difficult to repair. The danger is not only that individual elections might be won or lost on the basis of lies, but that the very idea of political truth is eroded. If any clip can be dismissed as a fake, accountability becomes impossible, and democracy begins to break down. That is why the recent reforms in New South Wales are so important and so timely.
New South Wales has decided to act before the damage is done. The next state election is scheduled for 13 March 2027, and on 7 April 2026 the Parliament amended the Electoral Act 2017 (NSW) to introduce two new offences—sections 189A and 189B—specifically targeting the publication and distribution of digitally generated electoral material during regulated election periods. In plain terms, when the official election campaign is underway, publishing political content that has been artificially generated or altered may be illegal unless specific conditions are met. The phrase “regulated election period” is central to the design. This is the window in which electoral regulation is at its most intense: rules about disclosure, advertising, and campaign expenditure apply, and voters are actively making up their minds. The Parliament has not told people to stop using artificial intelligence altogether. It has told them that if they do use AI to create realistic-looking electoral content, they must be transparent about it. Significantly, the amendments were passed well ahead of the election, not after a crisis. This gives candidates, political parties, media outlets, digital platforms, and ordinary citizens a chance to learn the rules and adjust their behaviour before the campaign begins. It also sends a clear signal that electoral law in Australia is paying attention to the new reality of synthetic media. Rather than waiting for a deepfake scandal to erupt and then scrambling for a response, New South Wales has chosen to put safeguards in place in advance. That is the mark of responsible and forward-looking legislation.
What exactly do sections 189A and 189B require? To appreciate the reforms, it helps to notice what they do not require. The law does not empower an electoral regulator, a judge, or a tribunal to decide whether a political claim is true, false, or misleading. That kind of inquiry is a minefield. Deciding truth in a political advertisement is notoriously controversial, and in a free society, officials should be extremely careful about policing the content of political speech. The New South Wales reforms take a more focused approach. They zero in on the synthetic character of electoral material. The decisive question is not “Is this statement accurate?” but rather “Is this image, audio, or video artificially generated or altered?” The reforms fall into two broad categories. The first category effectively prohibits certain simulations of real people. If you create a video in which a genuine candidate appears to say something they never said, or in which a real public figure is made to engage in behaviour that never took place, and you publish it as electoral material during a regulated election period, you may be committing an offence. This is a direct attack on the most dangerous form of deepfake: realistic fake footage of real individuals being used to deceive, discredit, or smear them. The second category is more subtle. It deals with content that appears to depict real people, events, or places but has actually been digitally generated. Such content is not automatically prohibited; rather, the law requires that it be accompanied by a clear disclosure or label. In other words, if something looks real but is not, the audience has a right to be told. For example, an AI-generated image of a candidate shaking hands with a historical figure could be lawful as satire if it is clearly labelled as synthetic, but without a label it could be dangerously misleading. The goal is not to block creativity; it is to ensure that voters are not tricked into treating fiction as fact.
Why is this approach distinctive? Because it refuses to solve the problem by having the government decide which political claims are true. That would be a disastrous road to travel. In a democracy, the state does not get to define objective truth in political conversation; that role is properly played by the public, the press, and the free exchange of ideas. The New South Wales approach, by contrast, is built around a more objective and technical fact: whether material has been artificially manufactured. Courts and electoral authorities are far more comfortable deciding that a video has been altered than deciding that a candidate’s policy proposal is false. This makes the law more enforceable and more respectful of free speech. The disclosure model, in particular, is elegant. It resembles the rules that require advertisements to be honest, food products to be clearly labelled, and medicines to list their ingredients. It does not assume that voters are stupid or fragile. On the contrary, it assumes that they are capable of judging for themselves, as long as they are given the essential information that the content they are seeing is not real. In this way, the law reinforces a key democratic value: transparency. It also weakens the power of deepfakes, because a disclosed fake is far less potent. The moment people know that a video is synthetic, its ability to deceive evaporates. At the same time, by focusing on regulated election periods, the law avoids unnecessarily policing artistic expression, satire, or everyday conversation throughout the year. It targets its energy at the moment when manipulation is most dangerous: the election campaign itself. This balance between protection and freedom is perhaps the greatest strength of the reforms.
Nevertheless, no law is a silver bullet. One major challenge is the complexity of defining “digitally generated” and “altered.” Almost every image on social media has been filtered or edited in some way. If a campaign brightens a photo or smooths out a shadow, should that require a label? Probably not, but where exactly does the line lie? The law will need to be interpreted sensibly, and regulators will need to provide clear, practical guidance. Another challenge is enforcement. Deepfakes can be produced in another country and released through anonymous accounts, making it difficult to identify a culprit within New South Wales. A state law is powerful but cannot solve a global problem on its own. Social media platforms therefore have a crucial responsibility: they must be willing to flag, remove, or label AI-generated content quickly, especially during elections. Their algorithms and policies will determine whether legal protections can keep up with the speed of viral dishonesty. There is also a human dimension. Many ordinary people may share a deepfake without realising that it is synthetic. If the law is applied too harshly, it could chill public participation and discourage people from engaging in political debate altogether. For that reason, public education and media literacy are essential. Voters need to become more critical consumers of online content; campaigners need to know the rules; and platforms need clear standards. Laws work best when they are widely understood and accepted, not when they are imposed without context. The New South Wales reforms are an important first line of defence, but they must be supported by a broader culture of verification, honesty, and digital awareness. Without that culture, even the best legal framework will struggle to protect democracy from the corrosive effects of synthetic disinformation.
At its heart, the New South Wales response to deepfakes is about protecting the possibility of trust. When we cast a vote, we are betting on shared reality. We must be able to agree that certain facts are true, that certain events happened, and that certain words were actually spoken. We cannot single-handedly stop artificial intelligence from becoming more powerful, and we cannot pretend that disinformation will ever go away entirely. But we can decide that, within our election campaigns, deliberate deception through synthetic media will not be tolerated. The reforms establish clear norms: if you want to participate in electoral debate using realistic but fabricated images or voices, you must be transparent. If you cross the line into outright simulation of a real person in order to deceive voters, you will face prosecution. These rules may be adjusted in the future, and they will inevitably interact with difficult questions of free speech, privacy, and technological change. But the direction is right. The law treats the electorate as adults, capable of making up their minds when the facts are placed before them. It protects the common ground of shared reality, which is the foundation of democratic governance. It also sends a practical signal. Politicians, consultants, and digital media strategists will think twice before using deepfakes as a weapon, because they know that the law is watching. Ordinary citizens, meanwhile, may be encouraged to pause before sharing that shocking video, to ask where it came from, and to check whether it has been labelled as synthetic. In doing so, they become participants in the defence of democratic integrity, rather than passive victims of manipulation. The 2026 amendments are not the end of the story; in many ways, they are the beginning. But they offer a powerful example of how law, technology, and civic responsibility can come together. In the end, the fight against deepfakes is not simply a legal or technical battle. It is a test of whether we still believe in the power of truth and the possibility of honest democratic life—even in an age of astonishing digital illusion.