Can elections be protected from being deepfaked?

While recent deepfakes featuring some Bollywood actors shocked the country, election-bound states were also rocked by political deepfakes, including doctored clips from a popular TV quiz show, with the clear intention of deceiving voters and swaying their beliefs.

Manipulated photos, audio, and video snippets are not new in politics. To improve his image, even Mussolini circulated a doctored photo of himself. Over the last decade, social media has persistently disseminated a vast quantity of fake audio, videos, or images during elections around the globe. It’s now deepfakes’ turn to foray into the political sphere ahead of the pivotal 2024 elections in countries like the US, UK, and India, which might even determine the direction of geopolitics for decades to come. A disturbing story from the New Hampshire primary at the start of the US election season was an automatically generated robocall featuring President Biden’s voice, which had been used to dissuade Democrats from taking part in the primary.

The deepfake age of elections began in the early 2020s. One of its first applications was prior to the 2020 Delhi elections. Three years is a long enough time; AI-driven deepfakes now have the potential to create a “perfect storm” of misinformation, thanks to significantly advanced technology. Furthermore, anyone may create them with readily available software. Even professional assistance is supposedly offered for a tiny cost.
In what way can deepfakes influence elections? The general election of Slovakia in September, which many saw as a “test case” for the 2024 global elections, may serve as a trailer. Just two days before the election, a deepfake audio of the Progressive Slovakia party’s leader talking about rigging the election and doubling the price of beer went viral. Progressive Slovakia lost the election in a close contest. Were a significant number of electorates deceived by the deepfake audio? Who knows! The November election in Argentina happened in the realm of AI Hogwarts; AI was present everywhere. “Is Argentina the First AI Election?” was the title of an article run by the New York Times.

As AI-generated deepfakes are expanding wings, there’s every indication that they’ll be heavily utilised in the 2024 elections as well, with enhanced rigour, perhaps. Deepfakes with political overtones have already overtaken the US political landscape. An altered version of a TV interview featuring Democratic Senator Elizabeth Warren became well-circulated. An AI-generated ad featuring numerous hypothetical tragedies that would occur if Biden were re-elected was released by the Republican National Committee. When Ron DeSantis, the governor of Florida, declared his candidature for the Republican nomination, Donald Trump made fun of him by creating an AI-generated parody video of the announcement, which included the voices of Adolf Hitler, Elon Musk, and DeSantis. The DeSantis campaign also released an image of Trump embracing Anthony Fauci, whom Republicans utterly dislike for his role during the pandemic.

It makes sense why the American Association of Political Consultants issued a warning, stating that deepfake videos pose a “threat to democracy.” A few months ago, a deepfake image of Trump’s violent arrest went viral. Imagine the potential consequences if a fake photo like this becomes widely shared during the election season. A specific task force has been assembled by Biden’s 2024 campaign to prepare responses to deceptive AI-generated images and videos. However, how well might that work? Keep guessing.

The UK cybersecurity centre also stated that cyberattacks by adversarial nations and their proxies are proliferating and getting harder to track, and that deepfakes and other AI tools pose a threat to the next election. A purportedly fake AI-generated audio clip of Labour leader Keir Starmer recently went viral that seemed to capture him swearing at staffers.

Understandably, all around the world, policymakers are working hard to stop the exploitation of manipulated media, which has the potential to undermine elections and disenfranchise voters. However, as technology is on an ever-improving path, the challenge becomes more complex.

Justifiably, the Indian government is also concerned. More stringent laws and policies to stop the spread of deepfakes might be implemented, including addressing the detection and prevention of deepfakes, building a grievance and reporting mechanism, and raising awareness.

But will these be fool-proof? And how easy are they to implement, given the rapidly advancing technology? Deepfakes of candidates saying something undesirable in the final moments of a closely contested election might swing the outcome, as demonstrated in Slovakia. AI-generated images and videos are easy to make, and because of social media, they can go viral far faster than factcheckers can review and debunk them and take any legal action. There would already have been irreversible harm.

Deepfake is nothing but modern-day Photoshop technology. These pictures and advertisements have some similarities to the edited photos and videos, deceptive messaging, and robocalls that have long been a part of society. Disinformation operations just made a number of logistical obstacles easier to overcome and gave them new life thanks to AI and its deepfakes avatar.

One may wonder, though, how much deepfakes might actually affect an election. However, assessing that is a notoriously challenging undertaking. The most significant harm, however, may be the erosion of “trust,” which is the most crucial component of a democratic process. And that kind of dystopia might be looming in 2024 if the security sector is unable to demystify deepfakes’ potentially harmful use cases.

(The writer is a professor of statistics, Indian Statistical Institute, Kolkata)