How deepfakes undermine truth and threaten democracy | Danielle Citron

譯者: SF Huang 審譯者: 易帆 余

[This talk contains mature content]

〔本演說含有成人內容〕

Rana Ayyub is a journalist in India

拉娜‧阿尤布是印度的記者，

whose work has exposed government corruption

她的報導曾經揭露過政府腐敗

and human rights violations.

以及侵犯人權的事件。

And over the years,

這些年來，

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

她也習慣了工作 所引來的尖酸批評和爭議。

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

但這些都不足以讓她準備好 面對 2018 年四月所發生的事。

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

她第一次看到時， 正和一名朋友坐在咖啡廳：

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

她看到一個長達兩分二十秒 以她為主角的性行為影片，

And she couldn't believe her eyes.

她無法相信她的眼睛。

She had never made a sex video.

她從來沒有拍過性影片。

But unfortunately, thousands upon thousands of people

但不幸的是，成千上萬的人

would believe it was her.

都相信那是她。

I interviewed Ms. Ayyub about three months ago,

大約三個月前，我為了我所寫的

in connection with my book on sexual privacy.

一本關於性隱私的書而去訪問她。

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

我是法律教授、 律師和民權擁護者。

So it's incredibly frustrating knowing that right now,

當我知道現行的法律沒辦法協助她時，

law could do very little to help her.

我感到非常挫折。

And as we talked,

我們交談時，

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

她解釋說她應該要料到 會有假的性影片出現。

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

她說：「畢竟，性通常 被用來貶抑和羞辱女人，

especially minority women,

特別是少數的弱勢女性、

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

敢挑戰威權男性的少數女性，」

as she had in her work.

比如，她在工作中所扮演的角色。

The fake sex video went viral in 48 hours.

四十八小時內， 假造的性影片就被瘋傳。

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

她所有的線上帳戶都被 該影片的截圖給洗版，

with graphic rape and death threats

夾帶著繪聲繪影的強暴和殺害威脅，

and with slurs about her Muslim faith.

還汙辱她的穆斯林信仰。

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

線上貼文指出 她現在「有空」可以約炮。

And she was doxed,

她的個資被外洩，

which means that her home address and her cell phone number

她的住家地址和手機號碼，

were spread across the internet.

在網路上處處可見。

The video was shared more than 40,000 times.

該影片被分享了超過四萬次。

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

當一個人成為這種 網路暴民攻擊的目標，

the harm is profound.

所受到的傷害會極度嚴重。

Rana Ayyub's life was turned upside down.

拉娜‧阿尤布的人生 被搞得亂七八糟。

For weeks, she could hardly eat or speak.

好幾週的時間她都幾乎 食不下嚥或無法說話。

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

她停止寫作並關閉了 所有的社群媒體帳戶，

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

對於記者來說，這是很難的抉擇。

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

她很怕踏出家門。

What if the posters made good on their threats?

萬一貼文的人履行了 他們的威脅怎麼辦？

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

聯合國人權理事會確認她並沒有發瘋。

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

理事會發佈了一份公開聲明， 說他們擔心她的安全。

What Rana Ayyub faced was a deepfake:

拉娜‧阿尤布面對的就是深偽技術：

machine-learning technology

它是一種機器學習科技，

that manipulates or fabricates audio and video recordings

能操弄或偽造錄音和錄影，

to show people doing and saying things

偽造的影音呈現出當事人

that they never did or said.

從未做過或說過的內容。

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

深偽技術的產物 看起來很真實，但不是真相，

they're total falsehoods.

完全是假造的。

Although the technology is still developing in its sophistication,

雖然這項技術的精密度還在發展中，

it is widely available.

但它已經可廣泛取得。

Now, the most recent attention to deepfakes arose,

最近深偽技術的關注度之所以提升，

as so many things do online,

和線上許多其它內容一樣，

with pornography.

和色情內容有關。

In early 2018,

2018 年初，

someone posted a tool on Reddit

有人在 Reddit 上張貼了一項工具，

to allow users to insert faces into porn videos.

使用者可以將面孔放入色情影片當中。

And what followed was a cascade of fake porn videos

接踵而來的是 傾洩而出的偽造色情影片，

featuring people's favorite female celebrities.

主打大家最愛的女性名人。

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

現今，你可以上 YouTube 找到大量的教學影片，

with step-by-step instructions

內有逐步指示，

on how to make a deepfake on your desktop application.

教你如何用電腦的應用程式 來製作深偽影片。

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

可能很快我們就能用手機 來製作深偽影片了。

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

正是我們基本的人性弱點

and network tools

與網路工具的相互作用，

that can turn deepfakes into weapons.

使得深偽技術變成了武器。

So let me explain.

讓我解釋一下。

As human beings, we have a visceral reaction to audio and video.

我們人類對影音會直覺反應。

We believe they're true,

我們會相信它們是真的，

on the notion that of course you can believe

你認為當然能相信

what your eyes and ears are telling you.

你的眼睛和耳朵所告訴你的事情。

And it's that mechanism

而正是這個機制

that might undermine our shared sense of reality.

有可能削減了我們感知真實的靈敏度。

Although we believe deepfakes to be true, they're not.

雖然我們誤信深偽技術下的 產物是真的，但它們不是。

And we're attracted to the salacious, the provocative.

我們會被猥褻的、 刺激的內容物所吸引。

We tend to believe and to share information

我們傾向相信並轉貼分享

that's negative and novel.

這些負面新奇的資訊。

And researchers have found that online hoaxes spread 10 times faster

研究者發現網路惡作劇散播的速度，

比真實故事快十倍。

than accurate stories.

Now, we're also drawn to information

我們也會被和我們觀點

that aligns with our viewpoints.

相同的資訊所吸引。

Psychologists call that tendency "confirmation bias."

心理學家稱之為「確認偏差」的傾向。

And social media platforms supercharge that tendency,

社群媒體平台會強化你偏好的傾向，

by allowing us to instantly and widely share information

因為那些平台讓我們 能夠迅速、廣泛地分享

that accords with our viewpoints.

和我們觀點相同的資訊。

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

深偽技術有可能對 個人及社會造成嚴重的傷害。

So, imagine a deepfake

所以，想像一支深偽影片

that shows American soldiers in Afganistan burning a Koran.

內容是美國士兵在阿富汗燒毀可蘭經。

You can imagine that that deepfake would provoke violence

各位可以想像那種深偽影片

會挑起針對這些士兵的暴力反擊。

against those soldiers.

And what if the very next day

如果隔天

there's another deepfake that drops,

又出來了一支深偽影片，

that shows a well-known imam based in London

內容是在倫敦知名的伊斯蘭教領袖

praising the attack on those soldiers?

讚美那些對士兵的攻擊行為？

We might see violence and civil unrest,

我們可能會看到暴力和內亂，

not only in Afganistan and the United Kingdom,

不僅是在阿富汗及英國，

but across the globe.

還有全球各地。

And you might say to me,

各位可能會跟我說：

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

「拜託，丹妮爾，那太牽強了。」

But it's not.

但這並不牽強。

We've seen falsehoods spread

我們已經見過在 WhatsApp

on WhatsApp and other online message services

及其它網路訊息上散播的謠言，

lead to violence against ethnic minorities.

導致對非主流人群的暴力。

And that was just text --

那還只是文字訊息——

imagine if it were video.

想像一下，若是影片會如何。

Now, deepfakes have the potential to corrode the trust that we have

深偽技術極可能會損害

我們對民主制度的信任。

in democratic institutions.

So, imagine the night before an election.

想像一下選前之夜。

There's a deepfake showing one of the major party candidates

有一支深偽影片播放 一位主要政黨的候選人

gravely sick.

病得很重。

The deepfake could tip the election

那支深偽影片有可能影響選舉，

and shake our sense that elections are legitimate.

並動搖我們對選舉合法性的感受。

Imagine if the night before an initial public offering

想像一下，某間大型全球銀行

of a major global bank,

在首次公開募股的前一晚，

there was a deepfake showing the bank's CEO

出現一支深偽影片， 內容是該銀行的執行長

drunkenly spouting conspiracy theories.

醉醺醺地大放厥詞說著陰謀論。

The deepfake could tank the IPO,

這支深偽影片可能會 重創這首次公開募股，

and worse, shake our sense that financial markets are stable.

更糟糕的是，會動搖我們對 金融市場穩定的感受。

So deepfakes can exploit and magnify the deep distrust that we already have

所以，深偽技術能利用 並放大我們對政治人物、企業領袖

in politicians, business leaders and other influential leaders.

及其他具影響力之領導人之 已有的深層不信任感。

They find an audience primed to believe them.

深偽影片一開始就誤導了它的觀眾。

And the pursuit of truth is on the line as well.

而對真相的追尋也會受到危及。

Technologists expect that with advances in AI,

技術人員認為隨著人工智慧的進步，

soon it may be difficult if not impossible

未來想要分辨出真實影片 和偽造影片的差別

to tell the difference between a real video and a fake one.

可能會變得非常困難 甚至無法分辨。

So how can the truth emerge in a deepfake-ridden marketplace of ideas?

所以，在一個充斥著深偽影片 想法的市場中，真相如何能浮現？

Will we just proceed along the path of least resistance

我們會不會選擇走阻力最少的路，

and believe what we want to believe,

相信我們想要相信的，

truth be damned?

而不管真相為何呢？

And not only might we believe the fakery,

我們不僅可能會相信偽造品，

we might start disbelieving the truth.

我們還可能會開始質疑真相。

We've already seen people invoke the phenomenon of deepfakes

我們已經看過有人利用 深偽事件的表象，

to cast doubt on real evidence of their wrongdoing.

使他們錯誤行為的真實證據遭到質疑。

We've heard politicians say of audio of their disturbing comments,

我們也聽過政客對他們 不利消息的回應：

"Come on, that's fake news.

「拜託，那是假新聞。

You can't believe what your eyes and ears are telling you."

不要相信你眼睛看到 和耳朵聽到的東西。」

And it's that risk

這種風險，

that professor Robert Chesney and I call the "liar's dividend":

羅伯‧特切斯尼教授和我 稱之為「騙子的紅利」：

the risk that liars will invoke deepfakes

這風險在於騙子會透過深偽技術

to escape accountability for their wrongdoing.

來為自己惡行脫罪。

So we've got our work cut out for us, there's no doubt about it.

所以毫無疑問的， 我們有艱鉅的工作要完成。

And we're going to need a proactive solution

我們會需要科技公司、立法者

from tech companies, from lawmakers,

法律執行者以及媒體

law enforcers and the media.

主動積極地提出解決方案。

And we're going to need a healthy dose of societal resilience.

同時我們也會需要 社會復原力的健康處方。

So now, we're right now engaged in a very public conversation

當下，我們正在與科技公司

about the responsibility of tech companies.

進行關於社會責任的公開談話。

And my advice to social media platforms

我們對社群媒體平台的建議，

has been to change their terms of service and community guidelines

已經影響了他們的 服務條款和社群準則，

to ban deepfakes that cause harm.

禁止會造成傷害的深偽技術。

That determination, that's going to require human judgment,

這樣的決定，需要人性的判斷，

and it's expensive.

代價很高。

But we need human beings

但我們需要人類

to look at the content and context of a deepfake

來查看深偽產物的內容和背景，

to figure out if it is a harmful impersonation

以判定它是否是有害的偽造品，

or instead, if it's valuable satire, art or education.

還是具價值的諷刺作品、藝術或教育。

So now, what about the law?

那麼，法律呢？

Law is our educator.

法律是我們的老師。

It teaches us about what's harmful and what's wrong.

法律教導我們什麼 是有害的、什麼是錯的。

And it shapes behavior it deters by punishing perpetrators

法律會透過懲罰犯罪者 以及讓受害者獲得補償救助，

and securing remedies for victims.

以此來規範行為。

Right now, law is not up to the challenge of deepfakes.

現在，法律還無法面對 深偽技術的挑戰。

Across the globe,

全球各地

we lack well-tailored laws

都缺乏適當的法律

that would be designed to tackle digital impersonations

來處理會侵犯性隱私、

that invade sexual privacy,

毀損名譽、

that damage reputations

造成情緒痛苦的數位偽造行為。

and that cause emotional distress.

What happened to Rana Ayyub is increasingly commonplace.

發生在拉娜‧阿尤布 身上的事越來越常見。

Yet, when she went to law enforcement in Delhi,

但，當她去找德里的執法機關，

she was told nothing could be done.

她得到的回應是執法機關無能為力。

And the sad truth is that the same would be true

令人可悲的是，就算在美國和歐洲，

in the United States and in Europe.

結果也是一樣。

So we have a legal vacuum that needs to be filled.

我們的法律有漏洞，需要填補。

My colleague Dr. Mary Anne Franks and I are working with US lawmakers

我的同事瑪莉‧安‧妮弗蘭克斯博士 和我跟美國的立法者合作，

to devise legislation that would ban harmful digital impersonations

策劃立法以禁止那些會造成傷害

且等同於偷竊身分的數位偽造行為。

that are tantamount to identity theft.

And we've seen similar moves

我們已經看到在冰島、英國

in Iceland, the UK and Australia.

和澳洲都有類似的行動。

But of course, that's just a small piece of the regulatory puzzle.

但當然，那只是控管拼圖的 其中一小片。

Now, I know law is not a cure-all. Right?

我知道法律並非萬靈丹，對吧？

It's a blunt instrument.

法律是一種鈍器。

And we've got to use it wisely.

我們得要明智地運用它。

It also has some practical impediments.

實際執法上也常遇到障礙。

You can't leverage law against people you can't identify and find.

法律對你辨識不出、 定位不了的人無能為力。

And if a perpetrator lives outside the country

如果犯罪者住在國外，

where a victim lives,

和受害者不同國家，

then you may not be able to insist

那麼你可能無法堅持

that the perpetrator come into local courts

要犯罪者到當地法庭去面對司法。

to face justice.

And so we're going to need a coordinated international response.

所以我們需要協調出 一致性的國際回應方式。

Education has to be part of our response as well.

我們的回應也得要包含教育。

Law enforcers are not going to enforce laws

執法者無法執行

they don't know about

他們不懂的法律，

and proffer problems they don't understand.

也無法提出他們不了解的問題。

In my research on cyberstalking,

在我的網路跟蹤研究中，

I found that law enforcement lacked the training

我發現執法人員缺乏訓練，

to understand the laws available to them

不了解他們能使用的法律有哪些，

and the problem of online abuse.

也不了解線上霸凌行為的問題。

And so often they told victims,

所以，他們通常會告訴受害者：

"Just turn your computer off. Ignore it. It'll go away."

「把你的電腦關機。 不要理它。就沒事了。」

And we saw that in Rana Ayyub's case.

在拉娜‧阿尤布的案例中就是如此。

She was told, "Come on, you're making such a big deal about this.

她得到的回應是： 「拜託，妳太大驚小怪了。

It's boys being boys."

男孩子都是這樣子的。」

And so we need to pair new legislation with efforts at training.

所以我們在新立法時 還要搭配對人員的培訓。

And education has to be aimed on the media as well.

且媒體也需要被教育。

Journalists need educating about the phenomenon of deepfakes

記者需要了解深偽的現象，

so they don't amplify and spread them.

他們才不會去放大和散播它們。

And this is the part where we're all involved.

這也是和我們大家都有關的地方。

Each and every one of us needs educating.

我們每一個人都需要被教育。

We click, we share, we like, and we don't even think about it.

我們會點閱、分享、按讚， 做這些時卻不會去思考。

We need to do better.

我們必須做得更好。

We need far better radar for fakery.

我們需要有更好的 判讀能力來辨識真偽。

So as we're working through these solutions,

當我們正在致力於這些解決方案時，

there's going to be a lot of suffering to go around.

也會伴隨著許多的困難和痛苦。

Rana Ayyub is still wrestling with the fallout.

拉娜‧阿尤布 仍在和事件的餘波苦戰。

She still doesn't feel free to express herself on- and offline.

不論在線上或離線， 她仍然無法自在地表達自己。

And as she told me,

她告訴我，

she still feels like there are thousands of eyes on her naked body,

她仍然覺得好像有成千上萬的人 在盯著她的裸體看，

even though, intellectually, she knows it wasn't her body.

即使理智上她也知道 那並不是她的身體。

And she has frequent panic attacks,

她常常會有突發性的恐慌，

especially when someone she doesn't know tries to take her picture.

尤其是當她不認識的人 想要拍她的照片時。

"What if they're going to make another deepfake?" she thinks to herself.

「如果他們又要拿照片去做 深偽呢？」她會這麼想。

And so for the sake of individuals like Rana Ayyub

所以，為了像拉娜‧阿尤布這樣的人，

and the sake of our democracy,

以及為了我們的民主，

we need to do something right now.

我們現在就得要採取行動。

Thank you.

謝謝。

(Applause)

（掌聲）