How deepfakes undermine truth and threaten democracy

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.

[This talk contains mature content]

Rana Ayyub is a journalist in India

whose work has exposed
government corruption

and human rights violations.

And over the years,

she's gotten used to vitriol
and controversy around her work.

But none of it could have prepared her
for what she faced in April 2018.

She was sitting in a café with a friend
when she first saw it:

a two-minute, 20-second video
of her engaged in a sex act.

And she couldn't believe her eyes.

She had never made a sex video.

But unfortunately, thousands
upon thousands of people

would believe it was her.

I interviewed Ms. Ayyub
about three months ago,

in connection with my book
on sexual privacy.

I'm a law professor, lawyer
and civil rights advocate.

So it's incredibly frustrating
knowing that right now,

law could do very little to help her.

And as we talked,

she explained that she should have seen
the fake sex video coming.

She said, "After all, sex is so often used
to demean and to shame women,

especially minority women,

and especially minority women
who dare to challenge powerful men,"

as she had in her work.

The fake sex video went viral in 48 hours.

All of her online accounts were flooded
with screenshots of the video,

with graphic rape and death threats

and with slurs about her Muslim faith.

Online posts suggested that
she was "available" for sex.

And she was doxed,

which means that her home address
and her cell phone number

were spread across the internet.

The video was shared
more than 40,000 times.

Now, when someone is targeted
with this kind of cybermob attack,

the harm is profound.

Rana Ayyub's life was turned upside down.

For weeks, she could hardly eat or speak.

She stopped writing and closed
all of her social media accounts,

which is, you know, a tough thing to do
when you're a journalist.

And she was afraid to go outside
her family's home.

What if the posters
made good on their threats?

The UN Council on Human Rights
confirmed that she wasn't being crazy.

It issued a public statement saying
that they were worried about her safety.

What Rana Ayyub faced was a deepfake:

machine-learning technology

that manipulates or fabricates
audio and video recordings

to show people doing and saying things

that they never did or said.

Deepfakes appear authentic
and realistic, but they're not;

they're total falsehoods.

Although the technology
is still developing in its sophistication,

it is widely available.

Now, the most recent attention
to deepfakes arose,

as so many things do online,

with pornography.

In early 2018,

someone posted a tool on Reddit

to allow users to insert faces
into porn videos.

And what followed was a cascade
of fake porn videos

featuring people's favorite
female celebrities.

And today, you can go on YouTube
and pull up countless tutorials

with step-by-step instructions

on how to make a deepfake
on your desktop application.

And soon we may be even able
to make them on our cell phones.

Now, it's the interaction
of some of our most basic human frailties

and network tools

that can turn deepfakes into weapons.

So let me explain.

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 believe

what your eyes and ears are telling you.

And it's that mechanism

that 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 information

that's negative and novel.

And researchers have found that online
hoaxes spread 10 times faster

than accurate stories.

Now, we're also drawn to information

that 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 information

that accords with our viewpoints.

Now, deepfakes have the potential to cause
grave individual and societal harm.

So, imagine a deepfake

that shows American soldiers
in Afganistan burning a Koran.

You can imagine that that deepfake
would provoke violence

against those soldiers.

And what if the very next day

there's another deepfake that drops,

that shows a well-known imam
based in London

praising the attack on those soldiers?

We might see violence and civil unrest,

not only in Afganistan
and the United Kingdom,

but across the globe.

And you might say to me,

"Come on, Danielle, that's far-fetched."

But it's not.

We've seen falsehoods spread

on WhatsApp and other
online message services

lead to violence
against ethnic minorities.

And that was just text --

imagine if it were video.

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.

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.

Thank you.

(Applause)