Deep-Fakes' Rise From Porn to Politics

Imagine the following scenario: You are a candidate running for presidential elections and only days before the final vote, someone in the opposing party posts a video of you giving an interview and making derogatory comments about a minority group. The video goes viral in a matter of minutes after it’s been shared on social media. For the average viewer the video seems to be authentic, and it reinforces a series of latent prejudices about the minority group you referred to in the video. A more careful observer would find the video a bit suspicious, first of all because it’s completely out of character for you to make such comments, and second, because of the timing of the remarks themselves, just a day before the final vote. Moreover, an even more attentive viewer would note some discrepancies with the shadows in the background of the video and your uncharacteristic facial expressions.

At this stage, even if you come out with an official statement in which you argue that the video is fabricated and even if the mainstream media publish further evidence to support your claims, it’s already too late, you’ve lost the presidential elections. Enough people in the swing states saw the video and believed it was authentic to tip the balance of the vote in the favour of the other candidate, and the rest is history. Forget about a chance to contest the validity of the elections or to run for any public role in the future, because your reputation has been tarnished beyond repair. Obviously, the above is a worst-case scenario in which everything goes bad for you, but that doesn’t mean that if you are a candidate such a scenario is impossible. Actually, in 2020 such a scenario has never been more likely courtesy of deep-fakes, which have become much more sophisticated and ubiquitous.

So what are deep-fakes and how did they make their way from porn to politics? 

Sophisticated computer programs enable virtually anyone with a decent computer and an Internet connection to create realistic-looking photos and videos of people saying and doing things that they did not actually say or do. This allows the spread of misinformation that goes far beyond fabricated chat comments and faked articles, because the reputational damage that deep-fakes cause is much harder to mitigate. Moreover, the amount of deep-fake videos being uploaded online is growing rapidly. While at the beginning of 2019 there were 7,964 deep-fakes online, just a year later that figure had jumped to 14,678, and it continues to increase worryingly (source). What’s even more concerning is that the technology is improving at a rapid pace, and experts predict that deep-fakes will soon be indistinguishable from real images, which would pose a range of political and social risks, according to a recent report published by The Brookings Institute:

• Distorting democratic discourse;
• Manipulating elections;
• Eroding trust in institutions;
• Weakening journalism;
• Exacerbating social divisions;
• Undermining public safety.

So far, the technology hasn’t been widely weaponised, and with a few exceptions, which I will mention below, it was mainly used for entertainment. For example, as of September 2019, 96% of deep-fake videos online were pornographic (source), and the handful of websites dedicated specifically to deep-fake pornography that emerged, collectively gathered hundreds of millions of views over the past year and a half. Deep-fake pornography involves the artificial synthesis of explicit videos that feature famous celebrities or personal contacts, and is almost always non-consensual. However, even though pornography will go down in history as the first use case to which deep-fake technology has been widely applied, it is of course not the only one. Recently, various media outlets have been using deep-fakes to educate people about the inherent threat of this new technology, but also to entertain viewers. Several deep-fake videos have gone viral recently, giving millions of people around the world their first taste of this exciting and at the same time alarming technology: 

• President Obama using obscene words to describe President Trump;
• Mark Zuckerberg admitting that Facebook’s true goal is to exploit its users;
• Bill Hader morphing into Arnold Schwarzenegger on a late-night talk show.

However, the aforementioned examples were created for the purpose of showcasing the capabilities of deep-fake technology and for entertainment. Elsewhere around the world, deep-fakes are beginning to cause mayhem as they spread to the political sphere:

• In Gabon, the military launched an ultimately unsuccessful coup after the release of an apparently fake video of leader Ali Bongo suggested that the President was no longer healthy enough to hold office;
• In Malaysia, a video purported to show the Economic Affairs Minister having sex has generated a considerable debate over whether the video was faked or not, which caused reputational damage for the Minister;
• In Belgium, a political group released a deep-fake of the Belgian Prime Minister giving a speech that linked the COVID-19 outbreak to environmental damage and called for drastic action on climate change. 

As can be seen from the examples above, the risk of deep-fakes influencing politics is no longer a hypothetical one, but a reality in many countries around the world, Therefore, we now stand at an inflection point since in the span of only a few years deep-fake videos have grown from an Internet oddity to a destructive political force, and so we must act rapidly before the technology gets out of hand and begins to pose a serious national security risk.

One potential solution for governments to defend against deep-fakes is to pass laws that make it illegal to create or spread them. However, this approach poses a series of challenges both constitutional and practical in nature, and thus it hasn't been successfully used so far. The main hurdle is the same one as with fake news, which is that by passing down laws to proscribe online content, particularly political content, a government risks infringing on the freedom of speech of its citizens and influencing the political discourse to a certain degree. Therefore, the only feasible solution, at least in the short-term, is for tech companies to take more rigours action to limit the spread of harmful deep-fakes before they do any damage. Even though many citizens might feel uncomfortable to entrust private companies with this, ultimately the key is in their hands, since deep-fake videos are spreading on their platforms, and so they should be responsible for any content posted on their platforms by third parties. But this approach raises some very complex issues concerning free speech and censorship.

A series of startups have emerged (eg. Deeptrace, Truepic, etc.) that produce software to defend against deep-fake videos, and even though their sophisticated systems are promising, it is very unlikely that they will be able to prevent the spread of deep-fakes in the long term. The AI systems designed to root out deep-fake videos can’t keep up with the bad actors that are employing other AI programs to produce and constantly improve them at a rapid pace. It’s an arms race in which the bad guys seem to have a head start, and if they’re left unhindered it is likely that they will win, undermining democracy and our trust in each other. Thus, in a world polluted by fake news citizens’ ability to agree on what is true and what is not is more precarious than ever, and deep-fake videos threaten to accelerate this trend. Consequently, the 2020 elections can either mark the beginning of the end for democracy as we know it or a once in a generation opportunity to rein in deep-fakes before it’s too late. Will we enter a dark age of deep-fakes or an age based on preserving the truth?

Further reading:

• Deepfakes - Believe at Your Own Risk (The New York Times);
• How Deepfakes Make Disinformation More Real Than Ever (Bloomberg);
• The Dark Side Of Deepfake Artificial Intelligence And Virtual Influencers (Forbes);
• I created my own deepfake—it took two weeks and cost $552 (Ars Technica);
• Google has released a giant database of deepfakes to help fight deepfakes (MIT Technology Review);
• What should newsrooms do about deepfakes? These three things, for starters (NiemanLab).