WEBVTT - TikTok Safety Feature Investigation

0:00:02.520 --> 0:00:07.000
<v Speaker 1>Bloomberg Audio Studios, Podcasts, Radio News.

0:00:08.160 --> 0:00:11.879
<v Speaker 2>You're listening to Bloomberg Business Week with Carol Masser and

0:00:11.960 --> 0:00:14.400
<v Speaker 2>tim Stenoveek on Bloomberg Radio.

0:00:14.840 --> 0:00:16.800
<v Speaker 1>So we do want to get to a Bloomberg Business

0:00:16.840 --> 0:00:20.400
<v Speaker 1>Week exclusive original reporting. It's on a story about a

0:00:20.400 --> 0:00:24.560
<v Speaker 1>confidential document that says kept a safety feature from millions

0:00:24.600 --> 0:00:28.880
<v Speaker 1>of TikTok users, about fifteen million by design as part

0:00:28.920 --> 0:00:32.680
<v Speaker 1>of an experiment. One of those users died by suicide.

0:00:32.720 --> 0:00:35.800
<v Speaker 1>There is a lot of careful reporting here to get

0:00:35.840 --> 0:00:38.320
<v Speaker 1>to the story by the award winning and recognize Bloomberg

0:00:38.360 --> 0:00:42.440
<v Speaker 1>News investigative reporter Olivia Carvill. She joins us here on set.

0:00:42.479 --> 0:00:46.159
<v Speaker 1>The story, by the way, is also Today's Bloomberg Big Take, Olivia.

0:00:46.240 --> 0:00:49.559
<v Speaker 1>Good to have you hear. This story is about a document.

0:00:49.920 --> 0:00:51.840
<v Speaker 1>What is this document? What does it tell us?

0:00:52.000 --> 0:00:54.400
<v Speaker 3>Thank you so much for having me on. So I

0:00:54.440 --> 0:00:57.400
<v Speaker 3>think it's important to have context around this document, and

0:00:57.480 --> 0:00:59.680
<v Speaker 3>for that we've got to go back in time. Early

0:00:59.720 --> 0:01:02.880
<v Speaker 3>Truthy twenty two, a sixteen year old boy from Long

0:01:02.920 --> 0:01:07.080
<v Speaker 3>Island took his own life. He died by suicide. After

0:01:07.160 --> 0:01:11.120
<v Speaker 3>this happened, his mum, who didn't understand when searching through

0:01:11.120 --> 0:01:13.959
<v Speaker 3>his social media accounts, trying to find a sense of

0:01:14.000 --> 0:01:17.080
<v Speaker 3>what had happened to him. When she found his TikTok account,

0:01:17.400 --> 0:01:20.319
<v Speaker 3>she was shocked by the fact that the algorithm was

0:01:20.360 --> 0:01:25.119
<v Speaker 3>sending him non stop distressing, depressing videos, many of them

0:01:25.319 --> 0:01:29.760
<v Speaker 3>encouraging suicide. A year later, I reach out to the

0:01:29.760 --> 0:01:32.479
<v Speaker 3>company because I'm writing a story about this young boy

0:01:32.520 --> 0:01:34.560
<v Speaker 3>and the way the algorithm was sending in so much

0:01:34.640 --> 0:01:37.840
<v Speaker 3>sad content, and I asked the company what had happened

0:01:37.920 --> 0:01:41.440
<v Speaker 3>to Chase NASCAR's account? Why is the algorithm sending him

0:01:41.640 --> 0:01:45.800
<v Speaker 3>all of these videos about suicide? That sparked an internal

0:01:45.840 --> 0:01:49.560
<v Speaker 3>review at TikTok. It's an extraordinary review where they do

0:01:49.640 --> 0:01:53.080
<v Speaker 3>a deep dive into his watch history, what he was

0:01:53.120 --> 0:01:56.639
<v Speaker 3>seeing in the weeks, days, and hours before he died.

0:01:57.040 --> 0:01:59.800
<v Speaker 3>And that review is what is contained within this confidential

0:01:59.840 --> 0:02:00.840
<v Speaker 3>talk document.

0:02:01.560 --> 0:02:02.720
<v Speaker 2>Tell us more about Chase.

0:02:03.120 --> 0:02:06.480
<v Speaker 3>Who is he? Chase was a sixteen year old boy

0:02:06.520 --> 0:02:08.799
<v Speaker 3>from Long Island. He was a high school junior, an

0:02:08.800 --> 0:02:12.000
<v Speaker 3>elite athlete, really good at swimming, really good at soccer.

0:02:12.040 --> 0:02:14.760
<v Speaker 3>He had a big group of friends. He had no

0:02:15.160 --> 0:02:18.640
<v Speaker 3>mental health issues. There was no depression, no anxiety in

0:02:18.639 --> 0:02:21.840
<v Speaker 3>his background. He was an honors student, there was no

0:02:22.120 --> 0:02:26.399
<v Speaker 3>warning shots that there was a concern with his own

0:02:26.400 --> 0:02:29.880
<v Speaker 3>mental health. His death was a mystery to his friends,

0:02:29.880 --> 0:02:33.440
<v Speaker 3>to his family. But what happened in January of twenty

0:02:33.639 --> 0:02:37.320
<v Speaker 3>twenty two is that this young teenage boy started sharing

0:02:37.520 --> 0:02:42.799
<v Speaker 3>very dark videos on TikTok with his friends, videos about suicide.

0:02:43.120 --> 0:02:45.920
<v Speaker 3>They didn't understand why he was sharing this content, and

0:02:45.960 --> 0:02:49.560
<v Speaker 3>it happened very quickly. And this document tells us why.

0:02:51.400 --> 0:02:52.360
<v Speaker 1>What does it tell us?

0:02:52.720 --> 0:02:56.480
<v Speaker 3>It tells us that this teenage boy's account, this very

0:02:56.520 --> 0:03:01.480
<v Speaker 3>normal teenage boy, was pushed into a filter bubble of

0:03:01.600 --> 0:03:05.040
<v Speaker 3>suicide and self harm content. That's like an online echo

0:03:05.160 --> 0:03:11.000
<v Speaker 3>chamber where he is repetatively recommended depressing sad videos, many

0:03:11.080 --> 0:03:14.520
<v Speaker 3>of them encouraging suicide. And it tells us why he

0:03:14.680 --> 0:03:16.280
<v Speaker 3>was pushed into that filter bubble.

0:03:16.360 --> 0:03:18.840
<v Speaker 1>So let's go there. Why was he pushed into this

0:03:18.960 --> 0:03:20.120
<v Speaker 1>so called filter bubble?

0:03:21.040 --> 0:03:23.040
<v Speaker 3>So again, if we go back in time in early

0:03:23.040 --> 0:03:27.320
<v Speaker 3>twenty twenty two, TikTok has realized that these filter bubbles

0:03:27.360 --> 0:03:30.760
<v Speaker 3>can be quite harmful for its users, and at this point,

0:03:30.800 --> 0:03:33.920
<v Speaker 3>TikTok is one hundred and fifty million American users. They

0:03:34.000 --> 0:03:36.920
<v Speaker 3>want to make the platform safer, so they design a

0:03:36.960 --> 0:03:41.240
<v Speaker 3>feature to break up repetitive videos about the same topic.

0:03:41.360 --> 0:03:44.480
<v Speaker 3>So rather than getting sad video, sad video, sad video,

0:03:44.720 --> 0:03:47.840
<v Speaker 3>you get sad video, funny video, happy video, silly video.

0:03:48.120 --> 0:03:50.720
<v Speaker 3>They want to break up those filter bubbles, so they

0:03:50.800 --> 0:03:53.600
<v Speaker 3>roll out this feature, which they call a filter bubble

0:03:53.680 --> 0:03:59.520
<v Speaker 3>prevention strategy, but they hold back ten percent of US users.

0:04:00.080 --> 0:04:02.480
<v Speaker 3>That's one of the most important figures in this document.

0:04:02.600 --> 0:04:08.440
<v Speaker 3>They withhold fifteen million Americans from this safeguard because they

0:04:08.440 --> 0:04:12.040
<v Speaker 3>want to test engagement. They want to understand how this

0:04:12.160 --> 0:04:14.520
<v Speaker 3>safety feature is going to impact how long you stay

0:04:14.520 --> 0:04:15.200
<v Speaker 3>logged on for.

0:04:15.440 --> 0:04:18.279
<v Speaker 1>So this group that they held back who didn't get

0:04:18.400 --> 0:04:21.239
<v Speaker 1>essentially the safety future, right, that was their control group.

0:04:21.520 --> 0:04:23.640
<v Speaker 1>That was their control So this was an experiment.

0:04:24.040 --> 0:04:26.720
<v Speaker 3>This was an experiment in their own words in this

0:04:26.800 --> 0:04:29.799
<v Speaker 3>document they refer to it as an experiment, and Chase

0:04:29.880 --> 0:04:31.919
<v Speaker 3>Neske was in that control group.

0:04:35.760 --> 0:04:38.040
<v Speaker 2>The go ahead, Carol, Well, one.

0:04:37.920 --> 0:04:40.720
<v Speaker 1>Thing I want to ask, because you and I've talked

0:04:40.720 --> 0:04:42.960
<v Speaker 1>about this story and I've read it and we've read it.

0:04:43.560 --> 0:04:46.320
<v Speaker 1>When they held it back in the document, TikTok says

0:04:46.360 --> 0:04:49.240
<v Speaker 1>they held it back. They held back the safety future

0:04:49.520 --> 0:04:52.279
<v Speaker 1>by design. Why are those two words important.

0:04:52.640 --> 0:04:55.960
<v Speaker 3>Those two words are so important, Carol, because right now

0:04:56.040 --> 0:05:00.600
<v Speaker 3>in US courts, there is a broad litigation ledging that

0:05:00.720 --> 0:05:06.159
<v Speaker 3>social media platforms are dangerous and addictive by design, that

0:05:06.240 --> 0:05:10.560
<v Speaker 3>these companies have created platforms that it engineered for addiction,

0:05:10.880 --> 0:05:13.240
<v Speaker 3>that what their main goal is is to keep you

0:05:13.360 --> 0:05:16.520
<v Speaker 3>logged on for as long as possible, regardless of how

0:05:16.520 --> 0:05:19.400
<v Speaker 3>that may impact your mental health. And so for the

0:05:19.480 --> 0:05:23.440
<v Speaker 3>first time, we're seeing a social media platform use that

0:05:23.560 --> 0:05:28.599
<v Speaker 3>same language that they experimented on millions of US users,

0:05:28.640 --> 0:05:32.640
<v Speaker 3>including teenagers, and that as part of that experiment, they

0:05:32.760 --> 0:05:36.080
<v Speaker 3>withheld a safety feature from one of those teenage users.

0:05:36.320 --> 0:05:42.479
<v Speaker 3>Quote by design, This was the company intentionally deliberately holding

0:05:42.520 --> 0:05:43.600
<v Speaker 3>back that safeguard.

0:05:44.040 --> 0:05:47.760
<v Speaker 1>It speaks to there is an awareness, right, that's right.

0:05:47.800 --> 0:05:48.680
<v Speaker 3>It speaks to intent.

0:05:49.560 --> 0:05:52.560
<v Speaker 2>I want to read the company's response that they sent

0:05:52.560 --> 0:05:54.479
<v Speaker 2>in you in response to questions about the document. A

0:05:54.520 --> 0:05:58.120
<v Speaker 2>TikTok spokesperson center written statement saying the company was quote

0:05:58.320 --> 0:06:00.960
<v Speaker 2>deeply committed to the safety and wealth being of users,

0:06:01.080 --> 0:06:04.640
<v Speaker 2>especially teams. Our hearts break for any family that experienced

0:06:04.640 --> 0:06:07.960
<v Speaker 2>a tragic loss, the spokesperson said, quote to help protect

0:06:07.960 --> 0:06:11.560
<v Speaker 2>our community, we continue to invest significantly in trust and safety,

0:06:11.880 --> 0:06:15.920
<v Speaker 2>including robust detection systems and dedicated enforcement teams that proactively

0:06:15.960 --> 0:06:18.760
<v Speaker 2>remove content that violates our community guidelines.

0:06:18.880 --> 0:06:21.479
<v Speaker 1>So, Olivia, that was the company's response in the last

0:06:21.760 --> 0:06:22.600
<v Speaker 1>couple of weeks.

0:06:22.920 --> 0:06:25.560
<v Speaker 3>That was the company's response this week to this story.

0:06:25.600 --> 0:06:29.159
<v Speaker 3>And it's important to point out TikTok has forty thousand

0:06:29.200 --> 0:06:32.039
<v Speaker 3>people working on their trust and safety team. They have

0:06:32.200 --> 0:06:36.599
<v Speaker 3>huge investments, forty thousand people working on trust and safety.

0:06:36.920 --> 0:06:39.520
<v Speaker 3>More than a billion dollars is spent trying to safeguard

0:06:39.640 --> 0:06:43.080
<v Speaker 3>this platform. So there are huge efforts underway to do this.

0:06:43.400 --> 0:06:46.520
<v Speaker 3>But I also think what this document shows us is

0:06:46.520 --> 0:06:50.160
<v Speaker 3>that sometimes these decisions are made too fast without thinking

0:06:50.200 --> 0:06:51.799
<v Speaker 3>about the impact on the user.

0:06:52.480 --> 0:06:54.080
<v Speaker 2>Can you talk about this and you answered this a

0:06:54.120 --> 0:06:56.279
<v Speaker 2>little bit before we read the statement, but the broader

0:06:56.320 --> 0:07:00.560
<v Speaker 2>trend of social media companies over the last couple of years,

0:07:01.480 --> 0:07:03.640
<v Speaker 2>I mean, you and the team did a great documentary

0:07:03.640 --> 0:07:07.960
<v Speaker 2>on this about the pressure that they've come under, not

0:07:08.080 --> 0:07:12.040
<v Speaker 2>just from lawmakers, but also from lawsuits that have been

0:07:12.040 --> 0:07:13.480
<v Speaker 2>filed on behalf of families.

0:07:13.800 --> 0:07:17.000
<v Speaker 3>Yeah, and I was mentioning the litigation earlier, So this

0:07:17.160 --> 0:07:20.520
<v Speaker 3>is thousands of lawsuits that have been filed right across

0:07:20.600 --> 0:07:21.080
<v Speaker 3>the country.

0:07:21.080 --> 0:07:22.800
<v Speaker 1>Can I just put a number, because then you put

0:07:22.840 --> 0:07:25.080
<v Speaker 1>a number on it, you said. More than four thousand

0:07:25.080 --> 0:07:29.080
<v Speaker 1>personal andury lawsuits have been filed accusing social media companies

0:07:29.120 --> 0:07:33.640
<v Speaker 1>of harming children's mental health. Many, though rate have been consolidated.

0:07:33.120 --> 0:07:36.400
<v Speaker 3>YEAP been consolidated into two tracks, one in state court

0:07:36.480 --> 0:07:39.960
<v Speaker 3>one in federal court. In addition to those four thousand lawsuits,

0:07:40.000 --> 0:07:44.040
<v Speaker 3>twelve hundred school districts have filed the same case against

0:07:44.040 --> 0:07:46.960
<v Speaker 3>the social media platforms, alleging that they are addictive and

0:07:47.000 --> 0:07:50.400
<v Speaker 3>that they have drained school resources on like mental health counseling.

0:07:50.760 --> 0:07:53.360
<v Speaker 3>And in addition to that, we're seeing forty two state

0:07:53.360 --> 0:07:57.480
<v Speaker 3>attorneys general file suit against the companies alleging the same thing.

0:07:57.800 --> 0:08:01.640
<v Speaker 3>So there is massive litigation unfolding across the US right now.

0:08:01.800 --> 0:08:05.000
<v Speaker 3>We saw the very first trial earlier this year where

0:08:05.120 --> 0:08:07.880
<v Speaker 3>Google and Meta were held accountable for harming the mental

0:08:07.880 --> 0:08:11.280
<v Speaker 3>health of a twenty year old woman, and documents like

0:08:11.400 --> 0:08:16.320
<v Speaker 3>this are surfacing right now because of that litigation. These

0:08:16.320 --> 0:08:20.000
<v Speaker 3>companies went through a two year discovery process and they

0:08:20.040 --> 0:08:24.440
<v Speaker 3>handed over more than six million internal documents to the lawyers.

0:08:24.840 --> 0:08:27.120
<v Speaker 3>So now we get a first kind of behind the

0:08:27.160 --> 0:08:30.160
<v Speaker 3>scenes look at what was really happening inside the world's

0:08:30.200 --> 0:08:35.400
<v Speaker 3>biggest social media platforms as they fought to grow their users.

0:08:35.480 --> 0:08:38.559
<v Speaker 3>And this is an incredibly competitive space. You know, the

0:08:38.920 --> 0:08:41.240
<v Speaker 3>main goal is to keep you logged on for as

0:08:41.320 --> 0:08:43.920
<v Speaker 3>long as possible, and they don't want to lose users

0:08:43.960 --> 0:08:47.640
<v Speaker 3>to other platforms, and so this is part of that.

0:08:47.960 --> 0:08:50.400
<v Speaker 1>I mean, that's the thing, right, It's about bringing more

0:08:50.480 --> 0:08:55.320
<v Speaker 1>users and then also set like kind of managing against safety,

0:08:55.440 --> 0:08:57.040
<v Speaker 1>like managing those two things.

0:08:57.160 --> 0:08:59.800
<v Speaker 3>And what it's called in the trust and safety space, Carol,

0:08:59.920 --> 0:09:02.800
<v Speaker 3>is it's a trade off, a trade off between user

0:09:02.880 --> 0:09:06.800
<v Speaker 3>safety and user growth. How do we keep the platform

0:09:07.280 --> 0:09:10.360
<v Speaker 3>growing as fast as it can but keep it as

0:09:10.400 --> 0:09:13.280
<v Speaker 3>safe as it can be for the users. And too

0:09:13.360 --> 0:09:17.200
<v Speaker 3>often the companies, according to these lawsuits and according to

0:09:17.240 --> 0:09:21.040
<v Speaker 3>previous reporting, are dialing up the user growth priority and

0:09:21.120 --> 0:09:23.360
<v Speaker 3>dialing back the user safety priority.

0:09:23.559 --> 0:09:25.520
<v Speaker 1>It's so important in this story and I highly recommend

0:09:25.520 --> 0:09:27.880
<v Speaker 1>people check it out also on the Bloomberg But you

0:09:27.920 --> 0:09:30.200
<v Speaker 1>can tell in the reporting you're very careful about laying

0:09:30.240 --> 0:09:32.600
<v Speaker 1>out the timeline. I just want to go back to

0:09:32.640 --> 0:09:35.080
<v Speaker 1>that for a moment, because there was a congressional hearing

0:09:35.559 --> 0:09:38.440
<v Speaker 1>back in March on March twenty third, twenty twenty three,

0:09:38.480 --> 0:09:43.439
<v Speaker 1>with TikTok's CEO answering questions from lawmakers. Is that that's

0:09:43.440 --> 0:09:44.720
<v Speaker 1>an important moment in time.

0:09:44.559 --> 0:09:47.600
<v Speaker 3>Too hugely important for this timeline. So if we think

0:09:47.640 --> 0:09:50.320
<v Speaker 3>about what's happening now that we have this document, we

0:09:50.400 --> 0:09:53.640
<v Speaker 3>know that inside the company they were doing this review

0:09:53.679 --> 0:09:56.400
<v Speaker 3>of Chase Nesker's account to try and understand what had

0:09:56.400 --> 0:09:59.200
<v Speaker 3>happened to his account and why he was receiving so

0:09:59.320 --> 0:10:03.280
<v Speaker 3>many videos about suicide. They figure out why, and that's

0:10:03.320 --> 0:10:07.520
<v Speaker 3>because he had that safeguard withheld from his account. And

0:10:07.760 --> 0:10:10.959
<v Speaker 3>that same month, that very same month, the CEO is

0:10:11.040 --> 0:10:15.360
<v Speaker 3>testifying before Congress answering questions about child safety. And I

0:10:15.440 --> 0:10:17.440
<v Speaker 3>do want to point to just one of the follow

0:10:17.520 --> 0:10:21.400
<v Speaker 3>up questions that they got from Congress. They were asked, specifically,

0:10:21.760 --> 0:10:25.240
<v Speaker 3>how many cases is TikTok aware of where users have

0:10:25.360 --> 0:10:30.800
<v Speaker 3>committed suicide after receiving sad, depressing or suicide promoting videos.

0:10:31.559 --> 0:10:34.280
<v Speaker 3>We know the company knew that it had sent those

0:10:34.320 --> 0:10:37.040
<v Speaker 3>kind of videos to Chase Nascar because of this document,

0:10:37.440 --> 0:10:41.000
<v Speaker 3>and this was the written response. TikTok does not believe

0:10:41.120 --> 0:10:44.679
<v Speaker 3>that use of the platform has caused users to commit suicide.

0:10:47.000 --> 0:10:49.880
<v Speaker 1>And again, what's key is timing in all of this,

0:10:49.920 --> 0:10:53.160
<v Speaker 1>which you lay out. As he testified two weeks after

0:10:53.200 --> 0:10:56.760
<v Speaker 1>that in depth review, he said, TikTok dead of Chase's account.

0:10:56.920 --> 0:10:59.320
<v Speaker 3>That's right that this was in the written responses to

0:10:59.360 --> 0:11:03.120
<v Speaker 3>Congress that came from the company rather than from TikTok CEO.

0:11:03.280 --> 0:11:05.280
<v Speaker 3>We're not aware of whether or not he was made

0:11:05.280 --> 0:11:09.000
<v Speaker 3>aware of this review, but the company responded to congressional

0:11:09.080 --> 0:11:11.480
<v Speaker 3>questions and put on the record that they didn't believe

0:11:11.520 --> 0:11:14.920
<v Speaker 3>that their platform had caused any users to commit suicide.

0:11:14.960 --> 0:11:18.240
<v Speaker 3>And this was said after this review was conducted.

0:11:18.040 --> 0:11:21.120
<v Speaker 1>As we said, laid out, and the dates and the

0:11:21.160 --> 0:11:23.640
<v Speaker 1>timeline are so important, as well as some of the

0:11:23.720 --> 0:11:26.079
<v Speaker 1>words that were said. Thank you so much. We really

0:11:26.120 --> 0:11:29.720
<v Speaker 1>appreciate it always. Bloomberg News investigative reporter Olivia Carville her

0:11:29.760 --> 0:11:32.320
<v Speaker 1>full reporting. It is Today's Bloomberg Big Take. It is

0:11:32.360 --> 0:11:34.880
<v Speaker 1>on the Bloomberg. It can also be found at Bloomberg

0:11:34.920 --> 0:11:35.440
<v Speaker 1>dot com.