At Reality Defender, we frequently encounter malicious deepfakes deployed to disrupt democratic elections around the world. Among other cases, we have helped political candidates and investigative authorities in Canada and Taiwan expose deepfake attempts aimed at tarnishing candidates and spreading disinformation among the electorate.
With deepfake technology advancing rapidly and major elections looming, experts universally acknowledge that deepfakes are poised to become potent tools of disinformation and chaos. Their proliferation threatens the integrity of democratic elections everywhere and deepens the rift between public institutions and individuals. The gravity of this risk cannot be overstated; malicious deepfakes are here to stay, and their influence could have serious consequences for our society if we fail to implement adequate safeguards.
Misleading Political Statements
Deepfakes can be used to create realistic audio and video of political figures making misleading or false statements or claims. With the help of such content spreading on social media, malicious actors can malign the reputation of candidates, attribute false policy aims and political opinions to public figures, create confusion among voters, and influence public opinion in ways detrimental to a healthy and fair democratic process.
Fake Endorsements and Attacks
Deepfakes can be deployed to fabricate fake candidate endorsements from influential individuals or organizations. Similarly, deepfakes can also be used for fake condemnation and criticism, influencing the electorate with manufactured assertions. Such content can spread rapidly and gain legitimacy before it is flagged and debunked by moderators.
Fabrication of Incriminating Content
Malicious actors can use deepfake technology to create audio, video, and images of political candidates, their staff, election officials, and other public figures caught in compromising situations that never occurred. These materials can be used to malign a person’s reputation, turn voters against them, engage in blackmail and extortion, or even initiate criminal investigations based on manipulated evidence.
Impersonation of Election Officials
With off-the-shelf software and a few seconds of video and/or audio, engineers of disinformation can create sophisticated fake media of election officials making denigrating claims about the electoral process—alleging untrue instances of voter fraud, undue influence, declaring certain polling locations closed or spreading disinformation about voter eligibility. Such chaos could have a major impact especially on Election Day, when disinformation can easily outpace corrections and content moderation, causing issues for voters and election officials on the ground.
Undermining Trust in Democratic Institutions
Deepfakes can be deployed to create realistic news announcements of fake election results, spread fabricated “proof” of voter fraud and ballot tampering, and discredit the integrity of journalists and news media by creating manipulated content of false or biased reporting of election results. Regardless of whether these deepfakes are easily debunked, we have already seen that the mere deployment of this kind of disinformation has an immediate impact on events on the ground, systemically eroding the public’s trust in democratic institutions and the social contract.
The threat of deepfakes to democratic elections is a clear and present danger that demands immediate attention and action from governments, tech companies, media organizations, and citizens alike. As the technology continues to advance and become more accessible, we must remain vigilant and proactive in developing robust detection methods, implementing strict content moderation policies, and educating the public about the risks of manipulated media. The integrity of our democratic processes and the trust in our institutions hang in the balance. It is up to all of us to work together to combat the spread of disinformation and ensure that our elections remain free, fair, and transparent. The future of our democracy depends on it.
