How deepfakes undermine truth and threaten democracy (Quest special)

(The use of deepfake technology to manipulate video and audio for malicious purposes — whether it’s to stoke violence or defame politicians and journalists — is becoming a real threat. As these tools become more accessible and their products more realistic, how will they shape what we believe about the world? In a portentous talk, law professor Danielle Citron reveals how deepfakes magnify our distrust — and suggests approaches to safeguarding the truth. )

As human beings, we have a visceral reaction to audio and video.We believe they’re true,on the notion that of course you can believewhat your eyes and ears are telling you.And it’s that mechanismthat might undermine our shared sense of reality.Although we believe deepfakes to be true, they’re not.And we’re attracted to the salacious, the provocative.We tend to believe and to share informationthat’s negative and novel.And researchers have found that online hoaxes spread 10 times fasterthan accurate stories.Now, we’re also drawn to informationthat aligns with our viewpoints.Psychologists call that tendency “confirmation bias.”And social media platforms supercharge that tendency,by allowing us to instantly and widely share informationthat accords with our viewpoints.

Now, deepfakes have the potential to cause grave individual and societal harm.So, imagine a deepfakethat shows American soldiers in Afganistan burning a Koran.You can imagine that that deepfake would provoke violenceagainst those soldiers.And what if the very next daythere’s another deepfake that drops,that shows a well-known imam based in Londonpraising the attack on those soldiers?We might see violence and civil unrest,not only in Afganistan and the United Kingdom,but across the globe.

Now, deepfakes have the potential to corrode the trust that we have in democratic institutions. So, imagine the night before an election. There’s a deepfake showing one of the major party candidates gravely sick. The deepfake could tip the election and shake our sense that elections are legitimate. Imagine if the night before an initial public offering of a major global bank, there was a deepfake showing the bank’s CEO drunkenly spouting conspiracy theories. The deepfake could tank the IPO, and worse, shake our sense that financial markets are stable. 

So deepfakes can exploit and magnify the deep distrust that we already have in politicians, business leaders and other influential leaders. They find an audience primed to believe them. And the pursuit of truth is on the line as well. Technologists expect that with advances in AI, soon it may be difficult if not impossible to tell the difference between a real video and a fake one. 

So how can the truth emerge in a deepfake-ridden marketplace of ideas? Will we just proceed along the path of least resistance and believe what we want to believe, truth be damned? And not only might we believe the fakery, we might start disbelieving the truth. We’ve already seen people invoke the phenomenon of deepfakes to cast doubt on real evidence of their wrongdoing. We’ve heard politicians say of audio of their disturbing comments, “Come on, that’s fake news. You can’t believe what your eyes and ears are telling you.” And it’s that risk that professor Robert Chesney and I call the “liar’s dividend”: the risk that liars will invoke deepfakes to escape accountability for their wrongdoing. 

So we’ve got our work cut out for us, there’s no doubt about it. And we’re going to need a proactive solution from tech companies, from lawmakers, law enforcers and the media. And we’re going to need a healthy dose of societal resilience. So now, we’re right now engaged in a very public conversation about the responsibility of tech companies. And my advice to social media platforms has been to change their terms of service and community guidelines to ban deepfakes that cause harm. That determination, that’s going to require human judgment, and it’s expensive. But we need human beings to look at the content and context of a deepfake to figure out if it is a harmful impersonation or instead, if it’s valuable satire, art or education. 

So now, what about the law? Law is our educator. It teaches us about what’s harmful and what’s wrong. And it shapes behavior it deters by punishing perpetrators and securing remedies for victims. Right now, law is not up to the challenge of deepfakes. Across the globe, we lack well-tailored laws that would be designed to tackle digital impersonations that invade sexual privacy, that damage reputations and that cause emotional distress. What happened to Rana Ayyub is increasingly commonplace. Yet, when she went to law enforcement in Delhi, she was told nothing could be done. And the sad truth is that the same would be true in the United States and in Europe. 10:12

So we have a legal vacuum that needs to be filled. My colleague Dr. Mary Anne Franks and I are working with US lawmakers to devise legislation that would ban harmful digital impersonations that are tantamount to identity theft. And we’ve seen similar moves in Iceland, the UK and Australia. But of course, that’s just a small piece of the regulatory puzzle. 

Now, I know law is not a cure-all. Right? It’s a blunt instrument. And we’ve got to use it wisely. It also has some practical impediments. You can’t leverage law against people you can’t identify and find. And if a perpetrator lives outside the country where a victim lives, then you may not be able to insist that the perpetrator come into local courts to face justice. And so we’re going to need a coordinated international response. Education has to be part of our response as well. Law enforcers are not going to enforce laws they don’t know about and proffer problems they don’t understand. In my research on cyberstalking, I found that law enforcement lacked the training to understand the laws available to them and the problem of online abuse. And so often they told victims, “Just turn your computer off. Ignore it. It’ll go away.” And we saw that in Rana Ayyub’s case. She was told, “Come on, you’re making such a big deal about this. It’s boys being boys.” And so we need to pair new legislation with efforts at training. 

And education has to be aimed on the media as well. Journalists need educating about the phenomenon of deepfakes so they don’t amplify and spread them. And this is the part where we’re all involved. Each and every one of us needs educating. We click, we share, we like, and we don’t even think about it. We need to do better. We need far better radar for fakery. 

So as we’re working through these solutions, there’s going to be a lot of suffering to go around. Rana Ayyub is still wrestling with the fallout. She still doesn’t feel free to express herself on- and offline. And as she told me, she still feels like there are thousands of eyes on her naked body, even though, intellectually, she knows it wasn’t her body. And she has frequent panic attacks, especially when someone she doesn’t know tries to take her picture. “What if they’re going to make another deepfake?” she thinks to herself. And so for the sake of individuals like Rana Ayyub and the sake of our democracy, we need to do something right now.