YT YouTube2Text

YouTube2Text → Transcripts → Facebook whistleblower Frances Haugen on social media accountability and transparency

Facebook whistleblower Frances Haugen on social media accountability and transparency — Transcript

by Washington Post Live · 5,850 words · 875 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:09[Music]
0:10thank you
0:12[Music]
0:17the company's leadership knows how to
0:20make Facebook and Instagram safer but
0:22won't make the necessary changes because
0:24they have put their astronomical profits
0:26before people yesterday
0:29[Music]
0:39[Music]
0:41thank you
0:42[Music]
0:48the original book I set out to write was
0:51uh focused on the idea that people have
0:53a lot of reasons why they don't follow
0:54their hearts you know that they they're
0:56afraid of losing their jobs of losing
0:59whatever be impoverished losing their
1:01homes losing their spouses you know
1:02they're going to end up estranged from
1:04everything falling out of complacency
1:05and and I had all those experiences in
1:08my late 20s at the same time
1:10and I came back from it
1:12and I wanted to write a story about
1:15um that where do we find our strength
1:17like how do we stand up when we fall
1:18down because the hardest thing I did
1:20wasn't blow the whistle on Facebook but
1:22because I lived those experiences I was
1:24able to follow my heart and stand up in
1:26the truth
1:31hello and welcome to Washington Post
1:32Live I'm Kat zakreski a tech policy
1:35reporter here at the Washington Post and
1:38my guest today is the Facebook
1:39whistleblower who you might remember
1:41from Capitol Hill testimony in 2021
1:44Francis Haugen thank you so much for
1:46being with us here today at post live
1:49thank you for inviting me happy to be
1:51here
1:52and so you and I first met on a video
1:55call much like this almost two years ago
1:58when you were preparing to go public as
2:00the Facebook whistleblower can you tell
2:02me a little bit about your book The
2:05Power of one and what message you hope
2:07people take away from it
2:10you know we live in a world today where
2:14our information environment is more and
2:16more mediated by algorithms that you
2:20know are actually quite new you know
2:23even in the perspective of you know
2:25since section 230 was written you know
2:28our laws came from the 90s largely about
2:31technologies that were developed in the
2:3280s and now we spend a huge fraction of
2:37our Our Time online on systems where
2:39information is picked out using systems
2:41that weren't even really used in
2:44consumer products until the 19 19 the
2:472020s 20 teens 20 teens like 20 2011 I
2:51think it was the first year Facebook had
2:52an algorithmic feed
2:54um and one of the things I wanted was
2:56hoping that people would take away from
2:57it is I got to kind of see the Arc of a
3:00lot of these technological systems you
3:02know I worked on search quality at
3:03Google in the mid-2000s I worked on on
3:06ranking algorithms at Pinterest and in
3:08those very early seniors as like
3:11recommender systems were becoming a
3:13bigger and bigger part of our lives and
3:15then I got to be at face during the 2020
3:17election and see how how uh do the
3:20consequences of these systems play out
3:23when we don't have transparency and
3:25oversight and so I hope people take away
3:27from the book that we have ways of
3:29making the system safer we we have to
3:32act now that we are running out of time
3:34in terms of allowing the status quo to
3:37continue onwards and probably most
3:39importantly you know this is not a book
3:41just about social media it's about
3:43opaque technological systems so all the
3:46conversations we're having about
3:48generative AI about large language
3:50models those have lots of parallels to
3:54how we got in trouble with social media
3:55you know we can't inspect those systems
3:58we can't hold those systems accountable
4:00unless we pass laws that give the public
4:02the right to know
4:03and on that point about AI regulation
4:06President Biden is scheduled to meet
4:08later today with some consumer Advocates
4:10who you've worked for before you've
4:12worked with before Tristan Harris Jim
4:14steyer and others
4:16um I wanted to ask you I mean what
4:17should be the top priorities for the
4:19Biden Administration as they start to
4:21think about AI regulation
4:24I I know about that meeting
4:27I mean unfortunately in Europe uh right
4:29now so it it it
4:31um it it it life is about choices and
4:35you know prior commitments
4:37um I I think one of the major things we
4:39need to look at in terms of the risks of
4:42these AI systems is there are short-term
4:44high probability harms and there are
4:46longer term potentially higher uh
4:50magnitude harms but with ambiguous
4:52probabilities right so we we're spending
4:55a lot of time right now talking about
4:56hypothetical hypothetical existential
4:58risks so this is things like you know
5:01the AIS are going to wake up and come
5:02after us
5:04um and when we have very short-term very
5:08high probability risks like
5:10um information operations you know we
5:12already struggle with the role that
5:14social media has in our information
5:15environment once we have uh generative
5:19AIS you know a big part about how
5:22misinformation works is truth has to
5:25exist in a very narrow corridor you know
5:27there's there's one consensus reality
5:29you know misinformation gets to play
5:31across the entire Continuum of the
5:34information
5:34[Music]
5:35entire information continue
5:39um so they can have you know you can put
5:41an idea out there and it is exactly the
5:43catnip that like resonates with some
5:45bias you have in your brain when you
5:47move on to having large language models
5:49you know that AI can now produce 10 000
5:52variants a hundred thousand variants and
5:54find exactly the thing that is going to
5:57go most viral and that's that's a really
5:59serious problem you know we we one of
6:02the major ways that we caught
6:03information operations at Facebook uh
6:06was by looking at repetition you know
6:07the same messages are being shared over
6:09and over again from the same sources you
6:11know in a world where you can have way
6:14more variation a lot of those tools go
6:16away and that doesn't mean that we don't
6:18have options you know one of the major
6:20projects I worked on at Facebook
6:23um was something around influence
6:24modeling like the idea that real humans
6:26have relationships with each other that
6:28we influence each other and you actually
6:30pick that out in the data you know I
6:32write about it in the book
6:34it's very very hard to fake that kind of
6:38data to fake real human long-term
6:41interactions in a way that doesn't stand
6:44out for being an aberration
6:47um the thing though is tech companies
6:49don't have incentives for using
6:51technology like that you know every fake
6:53user you take off your platform makes
6:55your company slightly less valuable
6:57because you have fewer users and so the
6:59thing I would I hope that that panel
7:01will will talk about is we need to not
7:06sit in parallel sphere we need to be
7:09acting to make sure that we put in place
7:10incentives that encourage positive uses
7:13of these technology and decrease uh the
7:17uh the chance that these major problems
7:19like labor force displacement happen in
7:22the short term
7:23and you've played a central role in the
7:26Congressional debate about how to
7:27regulate social media we're expecting
7:30Senate Majority Leader Schumer to
7:31outline his plan for how Congress can
7:33regulate AI tomorrow but what lessons do
7:37you think lawmakers can learn from their
7:39attempts to regulate social media as
7:41they set their sights on AI
7:44great question
7:46you know there's a lot of parallels like
7:47I mentioned before you know this is a
7:49book not so much about social media it's
7:51about opaque technological systems so
7:54that's you know in contrast in
7:57oh
7:59when you run through a rainstorm right
8:01before you go on Zoom like that's you
8:03gotta get yourself cold
8:06um and and you know for for most of the
8:0920th century we had an industrial system
8:11that was largely inspectable you know we
8:13could go and put a sensor outside a
8:15factory until the pollution was coming
8:17out of that factory we could take a
8:19product and run tests on it and say oh
8:21there's lead in this product you know we
8:23could take a car and we could crash it
8:25and say this car is unsafe in the
8:26following ways you know the gas tank is
8:28misinstalled and so this car is at risk
8:31for exploding
8:32um the uh the thing about our Modern
8:35Times And A lot of these AIS fall right
8:38in this bucket we can't inspect them
8:40right that it introduces a real power
8:43imbalance when the only people who get
8:46to have you know get to grade the
8:48homework of these companies are the
8:49companies themselves you know they know
8:51everything runs on the data center it's
8:53out of view the only people who get to
8:55see behind the curtain are the tech
8:57companies themselves
8:59um that's a that's a problem because as
9:01we saw social media companies cut
9:03corners
9:05so the the most basic thing I always say
9:07is we need to have uh rights for
9:09researchers so you companies need to
9:12make publicly accessible research apis
9:16um we need to if you're if your AI is
9:19not a consistent experience for all
9:21users you need to give ways of doing
9:23research on that because a huge part of
9:25what happened with Facebook or in social
9:27media in general is that every person
9:29gets a different experience it's much
9:31harder to do even basic accountability
9:33work on Facebook versus say Google
9:37um uh I think we need a consumer Bill of
9:39Rights right like you should be able to
9:41know how things are personalized to you
9:43you should be able to know how ads are
9:45targeted at you a lot of people are
9:47surprised with the idea that in China
9:48you're allowed to know what ad ad
9:51targeting criteria are being used to
9:52Target you but you know we don't have
9:54the right to know that
9:56um I think there's some things where in
9:57where the the European Union is going in
9:59a good direction in terms of
10:00transparency about content that is
10:02generated using um AI you know if you're
10:05looking at an image if you're looking at
10:06text that was written by a computer you
10:09know you should know that you're not
10:10having experiences mediated by another
10:12human right and then the last thing is I
10:14think we need to have a conversation
10:16about education like I think one of the
10:18places where um we are at risk for
10:20really creating more of like a two-tier
10:23Society is
10:25you know I and right now we have you
10:28know challenges with things like you
10:30have a kid with behavioral problems it's
10:32very expensive to hire a teaching Aid to
10:35make sure that kid can be around other
10:37kids
10:39um it's much cheaper to put them in a
10:40headset and have them go hang out in a
10:42metaverse classroom
10:44um and I one of the things that I think
10:45is going to be a place where AI it could
10:48be potentially very socially disruptive
10:49and I haven't heard enough conversations
10:51about it is you know Mark Zuckerberg
10:54said we're investing in AI to to enhance
10:57the metaverse so that means is you can
11:00put on your headset and you can have
11:02what feels like an authentic
11:03relationship with this robot
11:06um you know that that has a potential to
11:09be even more socially isolating than
11:10social media is today and I worry that
11:13the people that we're going to push
11:14towards these kind of um
11:17uh reductive social experiences are
11:20going to be people who are already
11:21marginalized in society you know instead
11:23of building senior centers or making
11:24sure we have senior Outreach you know
11:26we'll put grandma in the metaverse
11:27instead of hiring that Aid to make sure
11:29that kid can participate in school
11:31um with other with other children uh
11:33we'll put them in a virtual classroom
11:34and so these are all things where we
11:37should be having the conversations now
11:38and not waiting 20 years to be like oh
11:40how'd that happen
11:42and you talked a lot during your
11:44experience whistleblowing about the
11:45broken incentives for social media and
11:48putting profit above the risk of hate
11:50when you talk about those risks around
11:52loneliness and how AI might exacerbate
11:55some of those issues how can we change
11:57the incentive so that doesn't become our
11:59new reality
12:01great question
12:03um the so I think there's a couple
12:04things one is look before we even get to
12:07just the loneliness question like what's
12:08the more generalized incentives question
12:11um you know one of the things that
12:12frustrates me is uh you know open AI has
12:15spent more money on safety than I think
12:17basically any of the other big AI firms
12:20right like look at Facebook they fire
12:22their their AI safety team and then open
12:24source their model right like that
12:26that's about as as as different ends of
12:29the Continuum in terms of increasing
12:31generalized risk to society right you
12:35know uh they opening I could have gone
12:38even bigger Head Start and they chose to
12:40spend more time trying to figure out how
12:41to do the safety side right Facebook uh
12:44by by basically facilitating
12:47proliferation you know they they
12:49increase the number of players in the
12:51space by you know 100 000 coders you
12:53know 50 000 coders because they lower
12:56the the bar to entry so much
12:59um you know right now there's no Market
13:02incentive for acting in a safer way like
13:05safety takes time safety takes money all
13:09those things are are competitive
13:10disadvantages unless you have a system
13:13that gives additional rewards for acting
13:15in a safe Manner and one of the things
13:17that I I agree with Sam Altman on is
13:19that if you don't have conversations
13:21around incentives
13:24you're going to be in a situation where
13:27um
13:27you know if I'll give you an example
13:29like the Fortune 500 could come out and
13:31say hey we as a group have decided we
13:34are only going to buy products and
13:37services from companies that comply with
13:39a certain Base Bar of safety
13:41um you know if you're willing to uh you
13:44know comply in a way that shows you're
13:45being rigorous we're willing to direct
13:47economic resources towards you if you're
13:50not acting in a way that meets that
13:51safety bar we go sell to someone else
13:54right um that's a way of incentivizing
13:57good behavior
13:59um that is really meaningful because you
14:00will end up accelerating those good
14:02actors in a way that is is
14:04transformative
14:05um but that doesn't happen accidentally
14:07you know you have to have proactive
14:09involvement from people across Society
14:12saying hey we don't want to repeat what
14:13happened with social media we have to
14:15get out ahead of this
14:17and so on that point because you bring
14:20up social media again I know you write
14:22in your book and talk in interviews
14:23about how on the one-year anniversary of
14:26going public as the Facebook
14:27whistleblower everyone was asking you
14:29know what was the biggest impact and and
14:31would you do this again
14:33um but now almost two years later as we
14:35see things like the digital Service Act
14:37coming into Force things like the
14:39surgeon general's warning on children
14:41and social media recently I mean what do
14:43you think um was the biggest game
14:45changer of the Facebook files
14:48well I think the thing that I'm most
14:51personally proud of is
14:52um I one of the other things the
14:53European Union passed was whistleblower
14:55protections so in terms of things that I
14:58think are systematic
15:00um decreases of risk for society you
15:03know we we need whistleblowers right
15:05where when you have an economy that is
15:07run more and more by opaque systems
15:11um the people on the inside become
15:13critical
15:14um mechanisms of Public Safety like when
15:17you look at what I forgot the name of
15:19the organization it's like it's like the
15:20you know American Auditors Association
15:22or something like that like Financial
15:23Auditors like they don't call
15:25whistleblowers whistleblowers they call
15:27them Sentinels right because they're
15:29like the first line of defense
15:31um and so that's like the thing I think
15:33uh um I got to meet the woman
15:36um who was the Enron whistleblower who
15:39did a huge amount of work around passing
15:40the Whistleblower laws in the United
15:42States
15:43um and I felt it I felt honored to be
15:45part of her lineage because I got
15:46whistleblower protections because of her
15:48and I look forward to one day meeting a
15:50whistleblower who got whistleblower
15:52protections under that scheme but the
15:54second thing is you know I I really want
15:57to contextualize what the surgeon
15:59general's warning means in The Advisory
16:01means in the United States
16:03there have only been like 10 15 of these
16:05advisories since the 60s right uh
16:08they're things like uh smoking causes
16:11cancer
16:12um seat belts save lives breastfeeding
16:14helps children
16:16um these things are things where when I
16:18say them people are like yeah duh like
16:20cigarettes we all know that
16:22um but in the time period before those
16:24advisories came out there was a great
16:25deal of ambiguity
16:27and uh one of the things that has
16:28happened historically when one of those
16:30advisories takes place is within two to
16:33three years after it usually some kind
16:35of action moves forward
16:36we don't know what that action will take
16:38will it be a class action situation will
16:41it be a law will it be you know who
16:43knows but usually Society begins making
16:46movement in the time after those
16:47advisors come out and so uh you know if
16:50you had asked me two years ago because I
16:52love Facebook almost exactly two years
16:54ago right so I love Facebook about
16:57um like two years and one month ago I
16:58left Facebook
16:59um I came out you know three months from
17:02now in the future
17:03um four months from now in the future
17:05um but uh if you'd asked me is the
17:08Surgeon General going to make an
17:10advisory in the next two years
17:12you know I I don't think I would have
17:14ever said yes right like we were just
17:16not in that place as a society
17:18um and so uh I'm I'm really cautiously
17:21optimistic about what's going to unfold
17:22in the next couple years
17:24and on that experience of whistleblowing
17:26we have an audience question from Julie
17:29in California who asks have you
17:31experienced significant trouble getting
17:33work in the tech field since becoming a
17:35whistleblower
17:37so I uh I think if I wanted to go back
17:40to being a data scientist I don't think
17:42it would be a problem like I've gotten
17:44um I would say on the order of like five
17:45or six job offers uh all from uh small
17:48social networks like social networks
17:50that are trying to move into this space
17:53um there's like one or two that if I
17:55thought we were in a better place like
17:57from a regulatory framework perspective
17:59I would totally go take that offer
18:01um uh just because like there's certain
18:03strategies for approaching safety like I
18:06I worked on influence modeling at
18:07Facebook like I find that whole Space
18:09super interesting
18:11um I and I I it's interesting like I've
18:14had conversations with
18:16um fairly senior up people
18:18um you know I'm not talking about
18:19Executives I'm talking about director
18:20level kind of people at you know
18:22companies that you know the names of
18:25um who have said you know people noticed
18:28that you did not demonize Facebook right
18:30that that you tried to focus on what are
18:33the lessons we can all learn from how
18:35Facebook went astray
18:37um you didn't uh you didn't really Point
18:39like other than other than Nick Clegg
18:41because you know I have opinions about
18:43Nick Clegg in general I try to be pretty
18:45generous so the Facebook employees
18:47um and
18:49the uh and and that even-handedness you
18:54know it I I think it comes off as pretty
18:56obvious to most people that I'm not
18:59a malicious person right I I I'm a
19:02midwesterner like I if I could go like
19:04if I if I didn't feel like we had a
19:06public public crisis
19:08um I would much rather be coding by the
19:10ocean right I would much rather have a
19:12remote job and just live in Puerto Rico
19:15um and so I don't think it's gonna be a
19:17problem over the long term when I
19:18eventually and like I've done this long
19:20enough I would like to go go back to
19:22getting to play with my computer all day
19:24um but uh for now this is what I'm
19:27working on
19:28and I wanted to ask you though a little
19:30bit about the personal toll of
19:32whistleblowing you write in the book
19:33about uh when this process was heating
19:37up and you sent a signal message to the
19:39person you had just started dating kind
19:41of you know addressing how risky this
19:44was getting and can you tell me a little
19:46bit about the impact that this process
19:48had on you and your family
19:50so the the story I told in the book so
19:53um I started dating the guy I eventually
19:56married right so uh he's actually here
19:58in Amsterdam with me like he's going out
20:00and getting dinner for us because you
20:01know I'm talking to you all and I have
20:03to do another one of these things in a
20:04couple hours
20:06um and we started dating like maybe
20:08three or four days before
20:11um I came out in 60 minutes
20:14um uh and uh the day after I came out in
20:1860 minutes uh we were sitting with a
20:20threat researcher like they were giving
20:21us kind of the briefing on you know who
20:23what are the people in the dark webs uh
20:26saying about you you know like should we
20:27get you you know private security you
20:29know that kind of thing and um we had
20:32not locked down my so we'd lock down my
20:35social media so it you know made it
20:37private you know generally tried to
20:39decrease my profile on the internet
20:41um we have for not thought to do it for
20:42my mom so my mom is very very active on
20:45the internet and so the trolls had gone
20:48and trolled through all of my mom's
20:51social media posts going back years and
20:52years and years and
20:55um I texted I texted my partner and I
20:57was like or like at that point he was my
20:59boyfriend of like five days and uh was
21:02like Hey like you know we'd had a
21:04conversation about how two of his
21:06friends had gotten married and two
21:08people who said they would never get
21:09married
21:10and in their vows they had if one day
21:13your path diverges her mind I hope you
21:16take it right like if your path to Joy
21:19if your path to Joy diverges your mind I
21:21hope one day you take it and and I've
21:23always really loved that sentiment
21:25because I think it's really about
21:26um valuing like your partner as an
21:29independent human being right and and I
21:31commented to him I was like hey you know
21:34you don't know what you've signed up for
21:35you know like we could have stalkers we
21:38could have Paparazzi show up God because
21:40I was I was he had been my roommate I
21:42was not just started dating him kind of
21:43thing
21:44um you know if at any point this is too
21:46much you know I'm not going to judge you
21:48if you have to bounce and he said the
21:51most delightful thing which I I still
21:53hold him to this day he said I'm not
21:55afraid of trolls I am a troll
21:58um and it's funny because when Elon Musk
22:00took over Twitter
22:01um uh he had he had a bunch of troll
22:05accounts because he likes to troll
22:06Chinese information operations you know
22:08like everyone needs a hobby that's what
22:10I boxes is
22:12um and maybe a month into after Elon
22:15took over all of his accounts got taken
22:17down because they should have like
22:18there's no reason why Twitter should
22:20have allowed those fake accounts to
22:21exist
22:22um but uh but but beyond that it's crazy
22:25none of the consequences that I feared
22:27happening happened
22:29you know like we never got Paparazzi and
22:32I don't know if that's partially because
22:33like I live in Puerto Rico and it's a
22:35little hard to get there you know we
22:36never had people show up and scream at
22:38me you know I have open DMS on Instagram
22:41and Twitter you know anyone can message
22:43me you know people don't harass me
22:46um my email is open on my my website you
22:49can drop me an email say hi no one you
22:52know I have one person who I'm pretty
22:54sure is a little mentally ill who sends
22:56me conspiracies on a regular basis but
22:58they're not you know they're not out to
22:59get me
23:00um I've been incredibly crowdly grateful
23:02that I have had a very easy whistleblade
23:05experience compared to what many people
23:07experience
23:09um and so I will I will never I will
23:11never complain about anything that's
23:12happened to me so far
23:14on that note I mean what advice would
23:16you give for other Tech employees who
23:18are thinking about blowing the whistle
23:20and coming forward like you
23:23so the thing I always remind people is
23:24you don't have to be visible right so
23:27for every whistleblower who you see like
23:29me there's the hundreds or a thousand
23:32whistleblowers who are very subtly
23:34working in the background
23:36um you know I came forward publicly as I
23:39describe in the book because like my
23:41lawyers were like hey you know you poked
23:43the bear like there's there hasn't been
23:46a whistleblown incident of this
23:47magnitude
23:48you know since the 18t whistleblower
23:51decades ago
23:53um you know you need to understand they
23:55know who you are and they can out you at
23:57any time you know so if you want
23:58Facebook to get to choose how to
24:00introduce you to the world you can stay
24:01in the closet but like you know you're
24:03rolling the dice
24:05um I you don't have to do that like most
24:08people are not going to whistle blow at
24:09that magnitude
24:11um uh you can directly give documents to
24:13Congress you can give documents to the
24:15SEC
24:17um there you don't have to be visible
24:19but at the same time you are one of the
24:21most vital lines of defense for the
24:24public you know for most of these
24:26technological systems people can't see
24:28how they're operating unless people like
24:30you come forward
24:32um and so I I hope you never find
24:34yourself in a position where you feel
24:35like uh people's lives rest on on
24:39information that you know that the
24:40public is is being that's being kept
24:43from the public
24:44um but the the the second thing is if
24:46you do find yourself in that situation
24:49I hope you find at least one person that
24:52you feel like you can be honest with
24:53because one of the things I didn't
24:55appreciate until I came out was that
24:57another way in which I had had a very
24:59easy whistleblowing experience was that
25:01I lived with my parents during covid
25:03that most people who become
25:06whistleblowers you know by the time they
25:08come forward they're they're really
25:10ground down you know they they live with
25:12a secret alone for that that where they
25:16fear it impacts people's lives for often
25:18long periods of time and and that is an
25:21emotional trauma
25:23um and uh because I lived with my
25:26parents like during the whole period
25:27where you know what I was hearing at
25:29Facebook was very different from like
25:31what my lived experience was
25:33um I didn't have to wonder am I the one
25:36that's delusional are they the ones
25:39being unreasonable
25:41um I could just talk to my parents and
25:42be like I saw this today like this
25:43doesn't seem right or you know my Dad
25:45could come in and say hey like you know
25:48if if you were the lowest level
25:49technician in my laboratory so he runs
25:52like the place where all the blood work
25:54is done in a hospital
25:57um you know if you're my lowest level
25:58technician there would be a sign in the
26:01elevator there'd be a sign in the
26:02bathroom there'd be a sign in your break
26:04room saying did you see something that
26:07endangered patient safety call this
26:09number someone will listen to you
26:10someone will take you seriously like you
26:13have a Harvard MBA and you don't know
26:15who to call
26:16right you're working on National
26:18Security issues and you don't know who
26:19to call
26:20um and I think because I had that
26:23grounding right like I had two smart
26:25people who I trusted the Judgment of who
26:27were like what you're seeing is
26:28unreasonable
26:30um it allowed me to act with more
26:31confidence than I would have otherwise
26:33had
26:35when I wanted to bring the conversation
26:37back to Facebook which has recently had
26:39some major financial problems has had a
26:41series of layoffs do you think that
26:43those cuts have intensified some of the
26:46problems that you exposed around content
26:48moderation
26:51a hundred percent so it's on a couple
26:53different levels so um so for just to
26:55refresh people
26:56um I because this is I I I'm amazed how
27:00there are certain things about me that
27:02are have still not penetrated large
27:04chunks of the internet or of society you
27:07know a poor complaint in my initial sign
27:10of testimony was that we were being told
27:12that content moderation kept us safe
27:15but content moderation at least as it's
27:18done today does not scale right so
27:21there's lots and lots of languages in
27:22the world
27:23um when you make safety systems that are
27:25about picking and choosing which ideas
27:26are good and which ones are bad you know
27:28you have to go and rebuild those systems
27:30over and over again Language by language
27:32Regional dialect by Regional dialect
27:35changing topicality all the time it's
27:37very expensive
27:39um it Facebook just doesn't do it in
27:41lots of languages
27:43um I in the year after I came forward
27:46they invested a lot more in safety
27:50um but in like they I think that on they
27:53said they were going to double the
27:54amount they spent on safety and I think
27:56they did
27:57um unfortunately Elon Musk came through
27:59and bought Twitter and showed that you
28:01could lose 75 of your employees and
28:04there would be no consequences
28:06so I don't want to wait into the debate
28:08about like you know pros and cons of
28:10Elon but one things he did that has had
28:12serious consequences across the tech
28:14industry is uh he showed that you know
28:18you could make a significantly less uh
28:21safe product you know it wasn't just
28:23content moderator content moderation
28:25policies he changed he fired the people
28:27who were looking for child Predators who
28:29were looking for information operations
28:31who were looking for people compromising
28:33these systems
28:35um and that that no because uh we only
28:40require companies to report their
28:42economic numbers so this is profits
28:45losses the expenses that went into
28:47generating those profits but we don't
28:49require them to report their social
28:51ledger so this is the costs and harms uh
28:54that happen because of their products
28:57you can steal from the social Side Of
28:59The Ledger you know not invest in safety
29:01as much not be as rigorous and because
29:04it's invisible
29:06all it does is it looks like your
29:08balance your economic balance sheet is
29:10more generous and and Mark Zuckerberg
29:12has come out and said like you know the
29:13reason he this is Facebook's year of
29:16efficiency and he fired lots of safety
29:18people including the AI safety team
29:21um and I is that that you know Elon Musk
29:25showed him you could do it you know he
29:27ripped the Band-Aid off operably
29:30um and I I really worry that it is an
29:33illustration of when we have critical
29:36societal systems that are opaque you
29:39know they operate off in a data center
29:40where we can't inspect them
29:42um
29:43we will keep seeing uh as long as those
29:46incentives remain the same as long as
29:48you don't have to report data on that
29:49social Side Of The Ledger you're going
29:52to continue to see an economic incentive
29:54to cut those corners
29:55and we have just about a minute left and
29:58you write in your book that every reader
30:00can be a part of the solution of raising
30:02awareness and fixing social media what
30:05do you want our Washington Post viewers
30:07today to take away
30:09hmm
30:11um so I I totally get I want I want to
30:13be sympathetic here
30:14um so I'm a data scientist when you say
30:16things to me like data set you know I
30:18get very excited because I would love to
30:20sit and play with that
30:21um and I know that sit there and be like
30:23Francis like I I just don't care about
30:25tech transparency like like if you talk
30:27to me about you know child Predators you
30:30talk to me about misinformation
30:31campaigns I care about those things but
30:33like transparency like what do you
30:34actually get for transparency
30:36um I hope one of the things that people
30:37take away from a book is that for almost
30:40any harm you care about on these systems
30:44we cannot collectively you know that
30:47means uh through Advertiser boycotts
30:50through divestment campaigns through
30:52consumer boycotts protests uh through
30:55class action litigation there's a lot of
30:58different mechanisms we can use to pull
31:00these systems back towards the common
31:02good but if we don't have transparency
31:05we leave the platforms and that includes
31:07things like big AI companies we leave
31:10them creating their own homework right
31:13the floor the floor of where we can
31:15begin to have a democratic conversation
31:16about how to move forward has to begin
31:18with transparency
31:20um and so that's that's the thing I hope
31:22people will take away is that we have a
31:25chance to unleash a whole new generation
31:27of Specialists a whole uh era of
31:31conversations about how we want to live
31:33online but we can only do that if we get
31:36to peel back the curtain
31:38well we're just about out of time so
31:40we'll have to leave it there Francis
31:41thank you so much for joining us today
31:43at Washington Post Live
31:45my pleasure happy to be here
31:48and thank you for all joining us from
31:50home if you would like to tune in for
31:52future Washington Post Live Events you
31:54can check them out at
31:55washingtonpostlive.com I'm Kat zakreski
31:58thanks again

About this transcript

This page contains the full transcript of Facebook whistleblower Frances Haugen on social media accountability and transparency by Washington Post Live, generated from the public captions YouTube serves with the video. The transcript has 5,850 words across 875 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.