1
00:00:01,280 --> 00:00:04,600
Speaker 1: Hi, Mostly Human listeners, It's Lori Siegel. Today, I have

2
00:00:04,760 --> 00:00:05,640
Speaker 1: a treat for you.

3
00:00:06,040 --> 00:00:09,200
Speaker 2: I'm sharing an episode of Kara Swisher's podcast On with

4
00:00:09,280 --> 00:00:12,080
Speaker 2: Kara Swisher. In it, Kara hosts a panel of deep

5
00:00:12,119 --> 00:00:16,120
Speaker 2: fake experts, including me, to discuss the threat, the current scope,

6
00:00:16,160 --> 00:00:19,440
Speaker 2: and the future of sexually explicit defake abuse. Now, if

7
00:00:19,440 --> 00:00:21,680
Speaker 2: you've been listening to our four part series Searching for

8
00:00:21,720 --> 00:00:24,880
Speaker 2: Mister deep Fakes, this is a perfect cap. If you're

9
00:00:24,920 --> 00:00:26,880
Speaker 2: new to the subject, it's a fantastic primmer.

10
00:00:27,160 --> 00:00:30,479
Speaker 1: Hope you enjoy.

11
00:00:32,080 --> 00:00:35,559
Speaker 3: Hey, folks, award of warning before we begin. Today's episode

12
00:00:35,560 --> 00:00:40,360
Speaker 3: discusses sexual violence, including rape, graphic sexual content, and suicide,

13
00:00:40,520 --> 00:00:42,560
Speaker 3: so you might want to keep this in mind when

14
00:00:42,600 --> 00:00:57,760
Speaker 3: you decide if and where to listen. Hi, everyone from

15
00:00:57,800 --> 00:01:00,279
Speaker 3: New York Magazine and the Vox Media podcast Now Work.

16
00:01:00,520 --> 00:01:03,280
Speaker 3: This is on with Kara Swisher and I'm Kara Swisher.

17
00:01:03,600 --> 00:01:06,720
Speaker 3: My guests today are attorney Kerry Goldberg, tech journalists and

18
00:01:06,800 --> 00:01:11,080
Speaker 3: CEO of Mostly Human Media, Lori Siegel, and VS. Supermanian,

19
00:01:11,360 --> 00:01:15,759
Speaker 3: the Walter P. Murphy, Professor of Computer Science at Northwestern University.

20
00:01:16,160 --> 00:01:19,959
Speaker 3: We're talking about non consensual deep fake pornography or AI

21
00:01:20,040 --> 00:01:23,760
Speaker 3: generated explicit videos and images that use real people's likenesses

22
00:01:24,040 --> 00:01:27,400
Speaker 3: without their consent. Celebrities were often the target of early

23
00:01:27,480 --> 00:01:30,880
Speaker 3: pornographic deep fakes. Now, according to one survey, one in

24
00:01:30,959 --> 00:01:35,200
Speaker 3: eighteen's personally knows someone who has been targeted. Nearly all

25
00:01:35,280 --> 00:01:39,640
Speaker 3: deep fake porn depicts women. Despite laws and moderation efforts,

26
00:01:39,640 --> 00:01:43,680
Speaker 3: pornographic deep fake websites and so called newdify apps continue

27
00:01:43,720 --> 00:01:47,200
Speaker 3: to multiply, and a recent analysis from Wired found that

28
00:01:47,240 --> 00:01:51,360
Speaker 3: Elon musks rock chatbot is still being used to generate

29
00:01:51,440 --> 00:01:55,560
Speaker 3: and host non consensual deep fakes. I am not surprised

30
00:01:55,640 --> 00:01:59,240
Speaker 3: in the least. Carrie Goldberg is a victim's rights attorney.

31
00:01:59,280 --> 00:02:03,320
Speaker 3: Her law firm specializes in fighting Internet abuse and revenge. Porn.

32
00:02:03,680 --> 00:02:07,559
Speaker 3: Journalist Lori Siegel recently released Searching for Mister Deep Fakes,

33
00:02:07,560 --> 00:02:12,200
Speaker 3: an investigative docuseries created for TikTok and partnership with Paris Hilton.

34
00:02:12,600 --> 00:02:16,200
Speaker 3: It follows an investigation to unmask the anonymous operator behind

35
00:02:16,280 --> 00:02:20,080
Speaker 3: one of the Internet's most notorious deep fake porn platforms.

36
00:02:20,639 --> 00:02:24,760
Speaker 3: VS Subramanian is a top computer scientist and deep fake expert.

37
00:02:25,040 --> 00:02:27,880
Speaker 3: He led the team that developed the Global online deep

38
00:02:27,919 --> 00:02:32,399
Speaker 3: fake Detection system a free platform to help journalists detect

39
00:02:32,480 --> 00:02:35,320
Speaker 3: deep fakes. I think it's critically important to talk about

40
00:02:35,320 --> 00:02:38,120
Speaker 3: this issue. When I started my career, pornography was at

41
00:02:38,120 --> 00:02:41,000
Speaker 3: this heart of the problems with the early Internet and

42
00:02:41,080 --> 00:02:43,640
Speaker 3: the uses of it and the abuse of it was

43
00:02:44,080 --> 00:02:46,880
Speaker 3: built into the system at the start. A lot of

44
00:02:46,880 --> 00:02:48,560
Speaker 3: people were aware of it then and have had a

45
00:02:48,680 --> 00:02:51,280
Speaker 3: very hard time dealing with it even as technology has

46
00:02:51,320 --> 00:02:54,000
Speaker 3: gotten better and better at being more and more abusive.

47
00:02:54,040 --> 00:02:56,560
Speaker 3: So it's critical that we keep talking about it, we

48
00:02:56,680 --> 00:03:00,880
Speaker 3: criminalize this, and we start to hold platforms, big platforms

49
00:03:01,080 --> 00:03:05,240
Speaker 3: responsible for the stuff that they distribute. Our expert question

50
00:03:05,280 --> 00:03:08,760
Speaker 3: today comes from Wall Street Journal tech reporter Georgia Wells.

51
00:03:09,160 --> 00:03:11,840
Speaker 3: This is an important conversation and one we will keep having,

52
00:03:11,960 --> 00:03:15,160
Speaker 3: so stick around and one more thing before we get started.

53
00:03:15,160 --> 00:03:18,000
Speaker 3: If you're going to be in Washington, DC on July sixteenth,

54
00:03:18,000 --> 00:03:21,560
Speaker 3: please join us live at the Johns Hopkins University Bloomberg

55
00:03:21,600 --> 00:03:24,200
Speaker 3: Center for a taping up on I'll be talking to

56
00:03:24,240 --> 00:03:28,320
Speaker 3: Gina Rimando, Commerce Secretary under President Biden and former Rhode

57
00:03:28,320 --> 00:03:32,080
Speaker 3: Island governor about what AI means for the American workforce,

58
00:03:32,400 --> 00:03:35,160
Speaker 3: and as a special bonus before that, conversation. I'll be

59
00:03:35,200 --> 00:03:38,520
Speaker 3: speaking with Johns Hopkins University President Ron Daniels and the

60
00:03:38,600 --> 00:03:42,800
Speaker 3: University of Notre Dame President Father Robert Dowd about how

61
00:03:42,960 --> 00:03:46,760
Speaker 3: universities are approaching AI and workforce issues. You can get

62
00:03:46,800 --> 00:03:50,000
Speaker 3: tickets and learn more at boxmedia dot com. Slash Kara

63
00:03:50,120 --> 00:04:04,000
Speaker 3: Swisher Live, Carrie, Laurie and VS. Thanks for coming on on.

64
00:04:04,320 --> 00:04:05,000
Speaker 3: Could you be here?

65
00:04:05,320 --> 00:04:06,200
Speaker 4: Thrilled to be here?

66
00:04:06,560 --> 00:04:10,280
Speaker 3: Hello, Hello, So. I talked to experts back in January

67
00:04:10,360 --> 00:04:13,960
Speaker 3: about the proliferation of sexualized non consensual AI deep fikes,

68
00:04:14,000 --> 00:04:17,280
Speaker 3: including images of children on Elon muss x that were

69
00:04:17,360 --> 00:04:21,200
Speaker 3: generated by its groc chatbot. Obviously, AI created deep fake

70
00:04:21,279 --> 00:04:24,599
Speaker 3: porn hasn't gone away on Groc or elsewhere, and the

71
00:04:24,640 --> 00:04:27,400
Speaker 3: technology continues to improve. And like each of you to

72
00:04:27,440 --> 00:04:30,800
Speaker 3: explain the current scope of the problem and how pervasive

73
00:04:30,839 --> 00:04:35,280
Speaker 3: it is and who it's affecting, and how technically advanced

74
00:04:35,279 --> 00:04:37,200
Speaker 3: it is. We're going to get more into detail, So

75
00:04:37,279 --> 00:04:40,760
Speaker 3: go high level here, Carrie, We'll start with you, than Laurie,

76
00:04:40,960 --> 00:04:41,880
Speaker 3: than VS.

77
00:04:42,200 --> 00:04:42,320
Speaker 2: So.

78
00:04:42,880 --> 00:04:48,040
Speaker 5: I'm a plaintiff's attorney that sues tech companies, and generally

79
00:04:48,120 --> 00:04:52,479
Speaker 5: I see problems at the very early stages. So you know,

80
00:04:52,560 --> 00:04:55,960
Speaker 5: back probably seven years ago, we started getting situations where

81
00:04:56,160 --> 00:05:00,760
Speaker 5: celebrities and really high level content creators were deep faked.

82
00:05:01,200 --> 00:05:04,919
Speaker 5: And then over the last just i'd say year or so,

83
00:05:05,520 --> 00:05:09,000
Speaker 5: it started trickling down to middle schoolers getting deep faked.

84
00:05:09,160 --> 00:05:12,840
Speaker 5: But all the apps involved were these pretty obscure things.

85
00:05:12,880 --> 00:05:17,719
Speaker 5: But then everything totally changed when Elon started sending around,

86
00:05:17,760 --> 00:05:20,599
Speaker 5: you know, pictures of toasters in bikinis, and it just

87
00:05:21,000 --> 00:05:25,880
Speaker 5: unleashed this never before seen sort of situation where everyone

88
00:05:25,960 --> 00:05:29,200
Speaker 5: was able to create deep fakes and then immediately publish

89
00:05:29,240 --> 00:05:31,640
Speaker 5: them on the X platform.

90
00:05:31,640 --> 00:05:33,320
Speaker 3: So that's where you saw it, Laurie.

91
00:05:33,640 --> 00:05:36,000
Speaker 2: Yeah, when we started looking at this years ago, it

92
00:05:36,080 --> 00:05:41,160
Speaker 2: was very much focused on celebrities, oftentimes women in power politicians.

93
00:05:41,200 --> 00:05:42,599
Speaker 1: It seems like the first line of defense.

94
00:05:42,600 --> 00:05:45,080
Speaker 2: And I remember thinking at the time, oh, no, you know,

95
00:05:45,160 --> 00:05:47,040
Speaker 2: you might not care about these people. We should care

96
00:05:47,040 --> 00:05:49,240
Speaker 2: about these people. But what this means is this is

97
00:05:49,279 --> 00:05:52,160
Speaker 2: going to happen in our high schools, and unfortunately that's

98
00:05:52,160 --> 00:05:55,599
Speaker 2: what we're seeing happening. So I would just say it's rampant,

99
00:05:55,760 --> 00:06:00,359
Speaker 2: and it's happening in our communities. It's happening to women,

100
00:06:00,560 --> 00:06:03,800
Speaker 2: to all sorts of folks, and unfortunately it's something that

101
00:06:03,839 --> 00:06:06,640
Speaker 2: teenagers are learning and it's becoming commonplace.

102
00:06:06,920 --> 00:06:11,240
Speaker 4: Go ahead, yes, Kara, I'm a technologist, and so I've

103
00:06:11,279 --> 00:06:15,400
Speaker 4: spent a lot more time thinking about the technology over

104
00:06:15,440 --> 00:06:19,400
Speaker 4: the last you know, five six years ago, the ability

105
00:06:19,440 --> 00:06:23,760
Speaker 4: to create deep fakes was in its infancy, but over

106
00:06:23,760 --> 00:06:26,760
Speaker 4: the last five years, and in particular over the last

107
00:06:26,839 --> 00:06:29,640
Speaker 4: year or two, it's really gotten to a stage where

108
00:06:29,680 --> 00:06:33,240
Speaker 4: even experts like me on the detection of deep fakes

109
00:06:33,240 --> 00:06:38,160
Speaker 4: have difficulty eyeballing something, listening to something, and telling if

110
00:06:38,160 --> 00:06:40,520
Speaker 4: it's real or fake. So what we're seeing is something

111
00:06:40,560 --> 00:06:43,160
Speaker 4: where you really need humans and AI to come together

112
00:06:43,680 --> 00:06:47,200
Speaker 4: to work as some kind of partnership to try and

113
00:06:47,240 --> 00:06:50,640
Speaker 4: figure this out. And you know, until we can get

114
00:06:50,680 --> 00:06:52,920
Speaker 4: that right, I don't see an end in sight.

115
00:06:53,360 --> 00:06:56,640
Speaker 3: And how has the technology improved in the last year

116
00:06:57,279 --> 00:06:58,120
Speaker 3: Just ease of use.

117
00:06:58,720 --> 00:07:02,000
Speaker 4: Ease of use has always been there. The improvement has

118
00:07:02,000 --> 00:07:05,360
Speaker 4: been in the quality of the generator deep fakes. So today,

119
00:07:05,400 --> 00:07:09,280
Speaker 4: for example, we're seeing things like face swapping, so I

120
00:07:09,279 --> 00:07:15,280
Speaker 4: could record something of myself saying something thoroughly objectionable and

121
00:07:15,360 --> 00:07:18,680
Speaker 4: then put, you know, replace my face with that of

122
00:07:18,680 --> 00:07:22,040
Speaker 4: a celebrity in the past that was difficult to do.

123
00:07:22,560 --> 00:07:24,840
Speaker 4: There were things around the edges of the face that

124
00:07:24,920 --> 00:07:28,320
Speaker 4: would be noticeable. Today, much of that has been worked

125
00:07:28,320 --> 00:07:28,800
Speaker 4: out and.

126
00:07:28,880 --> 00:07:31,400
Speaker 3: Ease of use for normal people to do so correct.

127
00:07:31,440 --> 00:07:32,960
Speaker 4: Indeed, absolutely yeah.

128
00:07:33,000 --> 00:07:36,320
Speaker 3: So mister d Fakes launched in twenty eighteen and was

129
00:07:36,360 --> 00:07:39,440
Speaker 3: once again the largest deep fake porn site online, getting

130
00:07:39,480 --> 00:07:43,000
Speaker 3: seventeen million visits a month at its peak. Laurie, talk

131
00:07:43,000 --> 00:07:46,240
Speaker 3: about your investigation into mister deep Fakes and how it began.

132
00:07:47,000 --> 00:07:49,040
Speaker 2: I began a bit with an obsession, to be quite

133
00:07:49,040 --> 00:07:52,680
Speaker 2: honest with you. I remember it was twenty twenty two,

134
00:07:52,840 --> 00:07:56,080
Speaker 2: chatch Ept had launched. Everybody was talking about how AI

135
00:07:56,280 --> 00:07:58,160
Speaker 2: was going to just do these incredible things. It's going

136
00:07:58,200 --> 00:08:01,320
Speaker 2: to cure cancer, it's going to to you know, give

137
00:08:01,400 --> 00:08:02,480
Speaker 2: us all this free time.

138
00:08:02,600 --> 00:08:04,520
Speaker 1: And of course, I mean, car, you know this better

139
00:08:04,560 --> 00:08:04,920
Speaker 1: than anyone.

140
00:08:04,960 --> 00:08:06,920
Speaker 2: We're all kind of looking at this and saying, okay,

141
00:08:07,200 --> 00:08:09,400
Speaker 2: there are a lot of unintended consequences, and this is

142
00:08:09,440 --> 00:08:13,320
Speaker 2: moving very quickly without the correct guardrails. And someone told

143
00:08:13,360 --> 00:08:16,640
Speaker 2: me about mister deep Fakes. I will never forget going

144
00:08:16,640 --> 00:08:21,520
Speaker 2: to this site and opening it up and seeing thousands,

145
00:08:21,840 --> 00:08:26,600
Speaker 2: thousands of highly sexually graphic images, deep fakes of women

146
00:08:26,600 --> 00:08:29,480
Speaker 2: who did not consent, and they were violent in nature.

147
00:08:30,000 --> 00:08:33,080
Speaker 2: It was a website where a creator could say, you know,

148
00:08:33,440 --> 00:08:35,600
Speaker 2: could go and make money, and someone could go and

149
00:08:35,640 --> 00:08:38,439
Speaker 2: say I want this woman doing this, and they could

150
00:08:38,480 --> 00:08:38,920
Speaker 2: pay for it.

151
00:08:38,960 --> 00:08:41,119
Speaker 1: So it was beyond just a website. It was a platform.

152
00:08:41,400 --> 00:08:43,200
Speaker 2: And then what was even more concerning as we were

153
00:08:43,240 --> 00:08:46,400
Speaker 2: digging into it was, you know, there was everything from

154
00:08:46,600 --> 00:08:50,040
Speaker 2: training data of women's faces to create hyper realistic deep

155
00:08:50,120 --> 00:08:53,320
Speaker 2: fake pornography, and then there were forums where people would

156
00:08:53,360 --> 00:08:56,160
Speaker 2: exchange notes on here's the latest deep fake tools, here's

157
00:08:56,200 --> 00:08:58,720
Speaker 2: how you can do this. You can pay a premium

158
00:08:58,760 --> 00:09:01,720
Speaker 2: and get a deep fake training manual. And so my

159
00:09:01,800 --> 00:09:04,000
Speaker 2: initial thought was this isn't going to just be for

160
00:09:04,040 --> 00:09:06,520
Speaker 2: these high profile women and high profile I would say,

161
00:09:06,600 --> 00:09:08,760
Speaker 2: let's take it with a grain of salt. And I

162
00:09:08,800 --> 00:09:11,120
Speaker 2: remember thinking at the time, well, it's not going to

163
00:09:11,160 --> 00:09:13,760
Speaker 2: stop here. You know, this is just the beginning of it.

164
00:09:13,800 --> 00:09:15,760
Speaker 2: And by the way, you're looking at these forums and

165
00:09:15,800 --> 00:09:18,160
Speaker 2: you're seeing people say, is it wrong that I want

166
00:09:18,160 --> 00:09:20,360
Speaker 2: to do this to my sister in law? And oh,

167
00:09:20,440 --> 00:09:23,319
Speaker 2: this is so easy to do, and so it's almost

168
00:09:23,400 --> 00:09:27,320
Speaker 2: creating entertainment around deep fake abuse. And so as I

169
00:09:27,360 --> 00:09:30,080
Speaker 2: was looking at this, I thought, well, surely there's something

170
00:09:30,120 --> 00:09:33,920
Speaker 2: these women can do to help themselves, and the reality

171
00:09:34,000 --> 00:09:36,560
Speaker 2: was absolutely not. There was nothing they could do because

172
00:09:36,880 --> 00:09:39,559
Speaker 2: the man behind it was anonymous, and he'd been anonymous

173
00:09:39,600 --> 00:09:42,679
Speaker 2: for six years. And even so, even if he wasn't anonymous,

174
00:09:42,679 --> 00:09:45,400
Speaker 2: the laws hadn't caught up. And even when I went

175
00:09:45,400 --> 00:09:47,079
Speaker 2: out and started talking about this, and by the way,

176
00:09:47,080 --> 00:09:48,480
Speaker 2: you never want to be the person talking about deep

177
00:09:48,480 --> 00:09:50,520
Speaker 2: fake porn at the dinner table, but turns out I was.

178
00:09:50,880 --> 00:09:53,560
Speaker 2: People were like, well, the images aren't real that and

179
00:09:53,640 --> 00:09:57,200
Speaker 2: so it doesn't matter, and I'm sure everybody on here

180
00:09:57,280 --> 00:09:59,200
Speaker 2: is kind of just shaking their head. So we also

181
00:09:59,240 --> 00:10:02,120
Speaker 2: had a cultural problem. We needed people to realize that

182
00:10:02,120 --> 00:10:03,920
Speaker 2: these are so hyper realistic.

183
00:10:04,480 --> 00:10:05,920
Speaker 1: Actually this causes.

184
00:10:05,640 --> 00:10:08,520
Speaker 2: Deep, deep psychological abuse, and it can be used to

185
00:10:08,600 --> 00:10:11,840
Speaker 2: try to digitally abuse a woman, I say, a woman anyone,

186
00:10:12,280 --> 00:10:14,280
Speaker 2: And so that was what was alarming, and so we

187
00:10:14,360 --> 00:10:16,920
Speaker 2: decided we had to go find him and raise awareness.

188
00:10:16,960 --> 00:10:18,600
Speaker 1: So that was the beginning of the investigation.

189
00:10:19,240 --> 00:10:22,199
Speaker 3: Let's hear from Joanne Chu. She's an artist and actor.

190
00:10:22,360 --> 00:10:25,079
Speaker 3: One of the women featured in Laurie's investigation. She's talking

191
00:10:25,080 --> 00:10:27,600
Speaker 3: about the moment when she realized her image had been

192
00:10:27,720 --> 00:10:29,800
Speaker 3: used in porn on mister deep Fakes.

193
00:10:30,679 --> 00:10:33,040
Speaker 6: I was just bored one evening and I'm just gonna

194
00:10:33,040 --> 00:10:35,199
Speaker 6: google Google my name and see what comes up. I

195
00:10:35,320 --> 00:10:37,240
Speaker 6: kept scrolling down, and then I said.

196
00:10:37,040 --> 00:10:38,280
Speaker 4: What what what? What?

197
00:10:38,280 --> 00:10:38,520
Speaker 6: What?

198
00:10:38,520 --> 00:10:39,040
Speaker 4: What is this?

199
00:10:41,320 --> 00:10:45,439
Speaker 6: It's like my face on very uh disturbing and graphic

200
00:10:45,720 --> 00:10:51,760
Speaker 6: graphic content. It's you know, Joanne cho or Joanne Cheo.

201
00:10:52,679 --> 00:10:55,560
Speaker 6: In the beginning, it was just a few listings mister

202
00:10:55,559 --> 00:10:58,920
Speaker 6: deep fakes came up social media girls e Porner, and

203
00:10:58,960 --> 00:11:02,400
Speaker 6: then when I typed it in against it multiplied. It's

204
00:11:02,559 --> 00:11:06,560
Speaker 6: very much akin to physical assault. I have like less

205
00:11:06,559 --> 00:11:09,880
Speaker 6: than three thousand followers, but it doesn't matter if you're

206
00:11:09,960 --> 00:11:12,480
Speaker 6: famous or you're not. Just looking at mister dee fakes.

207
00:11:12,520 --> 00:11:15,720
Speaker 6: It's like they have a whole alphabetized database. At one point,

208
00:11:16,720 --> 00:11:18,920
Speaker 6: when I typed in my name, my stuff wasn't coming

209
00:11:18,960 --> 00:11:20,760
Speaker 6: up first anymore. It was all of this content.

210
00:11:21,280 --> 00:11:24,240
Speaker 3: So she compares to being physically assaulted. Carrie, does this

211
00:11:24,320 --> 00:11:26,680
Speaker 3: reflect what you've heard from your clients, because it sounds

212
00:11:26,720 --> 00:11:29,560
Speaker 3: like perpetrators are telling themselves that it isn't real harm.

213
00:11:30,160 --> 00:11:32,959
Speaker 5: Yeah, I remember this when I first started my firm

214
00:11:33,040 --> 00:11:36,520
Speaker 5: back in twenty fourteen, and it was everyone's attitude about

215
00:11:36,559 --> 00:11:39,640
Speaker 5: revenge porn and how you know, this is just an

216
00:11:39,720 --> 00:11:44,000
Speaker 5: online problem. And back then there were all these revenge

217
00:11:44,040 --> 00:11:47,839
Speaker 5: porn websites and people were searchable through Google, which would

218
00:11:47,960 --> 00:11:51,920
Speaker 5: lead people to the search engine results and it would

219
00:11:51,960 --> 00:11:55,640
Speaker 5: just be pages and pages of nude images. Like all

220
00:11:55,640 --> 00:11:58,880
Speaker 5: the harm was really really underestimated back then, and it

221
00:11:58,920 --> 00:12:04,280
Speaker 5: took a huge, huge amount of advocacy to change the

222
00:12:04,320 --> 00:12:07,880
Speaker 5: public opinion about the shame around being you know, nude

223
00:12:07,920 --> 00:12:11,840
Speaker 5: online and we're just we're seeing something really really similar.

224
00:12:12,000 --> 00:12:14,800
Speaker 5: The difference though, is that back you know, ten years ago,

225
00:12:15,400 --> 00:12:19,280
Speaker 5: women were being blamed for having shared an intimate image

226
00:12:19,280 --> 00:12:22,400
Speaker 5: in the first place, and with deep fakes, I don't

227
00:12:22,400 --> 00:12:25,880
Speaker 5: see that same kind of victim blaming because here no

228
00:12:25,960 --> 00:12:29,800
Speaker 5: one consent it to the image in the first place.

229
00:12:30,160 --> 00:12:33,920
Speaker 5: But there is still a lot of I think marginalization

230
00:12:34,240 --> 00:12:36,320
Speaker 5: and minimizing of the harm.

231
00:12:36,440 --> 00:12:38,559
Speaker 3: That it's not physical assault, that it's.

232
00:12:39,760 --> 00:12:43,360
Speaker 5: Yeah, I feel it is conceptually different than a physical assault,

233
00:12:43,480 --> 00:12:47,400
Speaker 5: but that doesn't mean that it's not still really really

234
00:12:47,679 --> 00:12:52,440
Speaker 5: harmful and humiliating for victims, especially like I see so

235
00:12:52,559 --> 00:12:57,240
Speaker 5: many young people who are like victims, and it's other

236
00:12:57,400 --> 00:12:59,600
Speaker 5: young people's doing it to them, you know, like where

237
00:12:59,600 --> 00:13:03,160
Speaker 5: it's on peer in junior high school and high school settings,

238
00:13:03,160 --> 00:13:05,320
Speaker 5: and a lot of times the offenders just think that

239
00:13:05,360 --> 00:13:09,680
Speaker 5: they're doing a prank or if you know, and always

240
00:13:09,760 --> 00:13:12,440
Speaker 5: the defense is, well, I barely shared it with anybody,

241
00:13:12,480 --> 00:13:15,720
Speaker 5: I didn't post it online. But these things find a

242
00:13:15,760 --> 00:13:17,000
Speaker 5: way to get published.

243
00:13:17,360 --> 00:13:20,600
Speaker 3: So deepig technology first to merge around twenty fifteen. VS

244
00:13:20,720 --> 00:13:24,240
Speaker 3: talk about the online communities that exist for sharing information

245
00:13:24,480 --> 00:13:28,400
Speaker 3: making so called newdified tools accessible and how has it

246
00:13:28,440 --> 00:13:31,199
Speaker 3: evolved over the past ten years in terms of how

247
00:13:31,240 --> 00:13:32,079
Speaker 3: it's used.

248
00:13:32,440 --> 00:13:36,120
Speaker 4: So Kara, there are a number of neutification AI tools.

249
00:13:36,679 --> 00:13:40,640
Speaker 4: So these are tools that are of various varieties. One,

250
00:13:40,760 --> 00:13:44,400
Speaker 4: as I mentioned earlier, is face swapping, where you've got

251
00:13:44,600 --> 00:13:49,680
Speaker 4: an individual who's nude performing some sex act and their

252
00:13:49,720 --> 00:13:52,760
Speaker 4: face has been replaced with the face of somebody like

253
00:13:52,960 --> 00:13:55,760
Speaker 4: miss Chew, who you alluded to a couple of minutes back.

254
00:13:56,000 --> 00:13:59,880
Speaker 4: So that's one class of techniques. Another is to use

255
00:14:00,800 --> 00:14:07,240
Speaker 4: massive databases of training data of existing nude individuals to

256
00:14:07,360 --> 00:14:10,800
Speaker 4: create and generate new nudes of people who are hyper

257
00:14:10,840 --> 00:14:16,120
Speaker 4: realistic who are performing acts which the real individual portrayed

258
00:14:16,480 --> 00:14:19,600
Speaker 4: never performed. So that's another class of techniques, and I

259
00:14:19,600 --> 00:14:21,840
Speaker 4: don't want to go into the gory details of how

260
00:14:21,840 --> 00:14:24,640
Speaker 4: that's done, but there are a number of techniques. These

261
00:14:24,640 --> 00:14:28,480
Speaker 4: techniques have evolved better and better every single year, and

262
00:14:28,520 --> 00:14:32,320
Speaker 4: so as a consequence, what we see is really a

263
00:14:32,440 --> 00:14:36,760
Speaker 4: huge number of websites that can disseminate such stuff, not

264
00:14:36,880 --> 00:14:41,560
Speaker 4: just the produced content, but also the tools that are

265
00:14:41,680 --> 00:14:45,120
Speaker 4: used to create and disseminate this stuff. And that's typically

266
00:14:45,160 --> 00:14:48,479
Speaker 4: done on the dark web through some kind of distribution

267
00:14:48,600 --> 00:14:49,840
Speaker 4: network on the dark web.

268
00:14:50,080 --> 00:14:53,080
Speaker 3: And what changed when Elon made it easy to do

269
00:14:53,160 --> 00:14:55,120
Speaker 3: so because for all eyes in the dark web, you

270
00:14:55,240 --> 00:14:58,280
Speaker 3: have to actually make an attempt to go get it right.

271
00:14:58,320 --> 00:15:00,240
Speaker 3: It's like go into a porn store or go going

272
00:15:00,280 --> 00:15:02,560
Speaker 3: to a porn movie or something back in the old days.

273
00:15:03,080 --> 00:15:05,160
Speaker 3: Talk a little bit about how that changed.

274
00:15:05,440 --> 00:15:09,440
Speaker 4: Well, I can't speak specifically to Elon and what he did,

275
00:15:09,560 --> 00:15:12,840
Speaker 4: but what I can say is that the more these

276
00:15:12,920 --> 00:15:19,239
Speaker 4: tools are available through normal channels, the more the audience

277
00:15:19,680 --> 00:15:23,360
Speaker 4: that gets access to them. And as soon as you

278
00:15:23,400 --> 00:15:26,720
Speaker 4: have a big audience, you have a bigger audience of

279
00:15:26,720 --> 00:15:32,840
Speaker 4: bad guys or as Carrie mentioned, just teenagers who don't

280
00:15:32,840 --> 00:15:37,160
Speaker 4: know better and haven't been taught better, possibly because their

281
00:15:37,200 --> 00:15:41,280
Speaker 4: teachers and their parents, however decent they may be, were

282
00:15:41,320 --> 00:15:44,080
Speaker 4: not aware of the threat, and by the time they

283
00:15:44,120 --> 00:15:46,120
Speaker 4: figured out it was too late.

284
00:15:46,720 --> 00:15:48,720
Speaker 3: Deep fakes and newifi apps are also a big problem

285
00:15:48,760 --> 00:15:51,720
Speaker 3: in schools. As we mentioned earlier, recent survey of roughly

286
00:15:51,720 --> 00:15:54,040
Speaker 3: five hundred and fifty US teens found that over half

287
00:15:54,320 --> 00:15:57,320
Speaker 3: had created at least one image using neuification tools, while

288
00:15:57,320 --> 00:16:00,120
Speaker 3: a third of the teens reported having their image created

289
00:16:00,120 --> 00:16:03,040
Speaker 3: and shared non consentually. I'm sure the numbers are higher. Actually,

290
00:16:03,400 --> 00:16:06,680
Speaker 3: Sites like mister deep Figs and apps like groc as

291
00:16:06,760 --> 00:16:09,880
Speaker 3: we know to have normalized creating nons consensual in some

292
00:16:09,960 --> 00:16:13,600
Speaker 3: cases violent sexual imagery, often of women. Laurie talk about

293
00:16:13,600 --> 00:16:15,480
Speaker 3: the consequences for young people.

294
00:16:15,760 --> 00:16:19,240
Speaker 2: Yeah, you know, I'll start. I always like to say,

295
00:16:19,280 --> 00:16:20,960
Speaker 2: like to try to make it a little bit personal.

296
00:16:21,040 --> 00:16:22,920
Speaker 2: I'll never forget, and I'm going to take it here

297
00:16:22,920 --> 00:16:25,240
Speaker 2: and I promise I'll get to young folks, but I'll

298
00:16:25,320 --> 00:16:27,520
Speaker 2: never forget. I think it was twenty twenty three when

299
00:16:27,560 --> 00:16:30,320
Speaker 2: a group of technologists approached me and say, can we

300
00:16:30,600 --> 00:16:32,360
Speaker 2: can we have AI attack you and see what it

301
00:16:32,400 --> 00:16:36,160
Speaker 2: could find. And because I am a crazy person, I said, sure,

302
00:16:36,320 --> 00:16:39,200
Speaker 2: let's see. And what they did was they had AI

303
00:16:39,320 --> 00:16:41,760
Speaker 2: come up with an attack against me. And it took

304
00:16:41,800 --> 00:16:45,000
Speaker 2: some true things, right, like I've interviewed Mark Zuckerberg before,

305
00:16:45,120 --> 00:16:46,720
Speaker 2: or I'm a technology journalist.

306
00:16:47,080 --> 00:16:49,560
Speaker 1: And then it ended up in front of an audience.

307
00:16:49,800 --> 00:16:52,600
Speaker 2: It ended up creating the created sexually explicit deep fix

308
00:16:52,600 --> 00:16:55,880
Speaker 2: of me. And I'll never forget because this I had consented.

309
00:16:55,920 --> 00:16:57,320
Speaker 2: I didn't realize they were going to do that, but

310
00:16:57,360 --> 00:16:59,880
Speaker 2: I had consented to this demo, and I'll never forget

311
00:17:00,080 --> 00:17:03,200
Speaker 2: looking at the audience look at these sexually explicit deep

312
00:17:03,200 --> 00:17:05,760
Speaker 2: fakes of me, and I could, as a person who

313
00:17:05,960 --> 00:17:08,760
Speaker 2: generally formed sentences for a living, I stopped being able

314
00:17:08,760 --> 00:17:09,600
Speaker 2: to form sentences.

315
00:17:09,760 --> 00:17:12,240
Speaker 1: I felt shame, I felt humiliation.

316
00:17:12,640 --> 00:17:14,840
Speaker 2: I felt like the world had seen me naked, to

317
00:17:14,880 --> 00:17:17,760
Speaker 2: be quite honest with you, and then I thought to myself, well,

318
00:17:17,760 --> 00:17:20,320
Speaker 2: if that's how I feel, and I know what this

319
00:17:20,440 --> 00:17:23,440
Speaker 2: technology is, imagine what that's going to be like for

320
00:17:23,560 --> 00:17:26,480
Speaker 2: an eleven year old girl, you know. And I think

321
00:17:26,560 --> 00:17:28,960
Speaker 2: that's what we started hearing. So I've spoken to a

322
00:17:28,960 --> 00:17:31,800
Speaker 2: lot of parents, a lot of teenagers, who've had this

323
00:17:31,880 --> 00:17:34,520
Speaker 2: happen to them. I spoke to one woman who said,

324
00:17:34,640 --> 00:17:37,000
Speaker 2: and you know, she's in her early twenties. She said

325
00:17:37,040 --> 00:17:39,159
Speaker 2: she after this was happening, and she went out and

326
00:17:39,160 --> 00:17:41,040
Speaker 2: she tried to get this stopped, and people said, there's

327
00:17:41,040 --> 00:17:43,520
Speaker 2: nothing you can do. They blamed her, very similar to

328
00:17:43,560 --> 00:17:45,840
Speaker 2: what Carrie is talking about. Back when we were talking

329
00:17:45,880 --> 00:17:49,159
Speaker 2: about non consensual pornography, people didn't quite understand it. She

330
00:17:49,240 --> 00:17:51,560
Speaker 2: told me she walked to the roof of her building

331
00:17:51,640 --> 00:17:54,399
Speaker 2: and she considered jumping off. And you talk to a

332
00:17:54,440 --> 00:17:57,760
Speaker 2: politician who describes this as digital rape. And so for me,

333
00:17:57,880 --> 00:17:59,879
Speaker 2: it was really confusing because I was looking at this

334
00:18:00,040 --> 00:18:04,080
Speaker 2: website that for me, this website represented this dystopian era

335
00:18:04,160 --> 00:18:07,320
Speaker 2: where we lose consent, in an ai era where you know,

336
00:18:07,400 --> 00:18:10,359
Speaker 2: our most intimate qualities can be weaponized against us. And

337
00:18:10,359 --> 00:18:13,359
Speaker 2: then externally we were hearing but it's not real. It

338
00:18:13,359 --> 00:18:17,720
Speaker 2: doesn't matter when it really impacts the way a young person,

339
00:18:18,359 --> 00:18:20,560
Speaker 2: any person walks through the world. We spoke to a

340
00:18:20,600 --> 00:18:23,479
Speaker 2: woman named Mollie who was a mother who had this

341
00:18:23,680 --> 00:18:27,639
Speaker 2: her her husband's best friend. Deep faked the whole neighborhood,

342
00:18:27,720 --> 00:18:32,120
Speaker 2: eighty women and the community and created it's sexually explicit

343
00:18:32,119 --> 00:18:35,080
Speaker 2: deep fakes, and she said it impacted how she would

344
00:18:35,119 --> 00:18:37,280
Speaker 2: go on the bus and look at people and wonder

345
00:18:37,680 --> 00:18:39,679
Speaker 2: if they'd seen this, and she had to go on

346
00:18:39,760 --> 00:18:42,320
Speaker 2: a she said, a porn hunt for her own face

347
00:18:42,480 --> 00:18:45,960
Speaker 2: on the web. And I think it really impacts from

348
00:18:46,000 --> 00:18:48,320
Speaker 2: a human perspective, how you walk through the world and

349
00:18:48,320 --> 00:18:49,120
Speaker 2: how you hold.

350
00:18:48,880 --> 00:18:51,520
Speaker 1: Yourself, and it can have reputational harm.

351
00:18:51,880 --> 00:18:54,720
Speaker 3: Right, So, Carrie, your law firm in your or ensues

352
00:18:54,840 --> 00:18:58,800
Speaker 3: quote psychos pervs, trolls and toxic tech. In January, you're

353
00:18:58,800 --> 00:19:01,399
Speaker 3: from file the lawsuit against say On behalf of Ashley

354
00:19:01,440 --> 00:19:04,639
Speaker 3: Saint Clair, a conservative influencer who had a child with

355
00:19:04,680 --> 00:19:07,400
Speaker 3: Elon Musk. She alleges that his company's Chapbot was used

356
00:19:07,400 --> 00:19:11,240
Speaker 3: to create and disseminate sexually explicit, non consensual images of her.

357
00:19:11,640 --> 00:19:14,479
Speaker 3: Talk about the lawsuit and how is it representative of

358
00:19:14,600 --> 00:19:16,960
Speaker 3: bigger legal issues around deep big porn.

359
00:19:17,200 --> 00:19:21,399
Speaker 5: When Grock suddenly had this new capability of being able

360
00:19:21,440 --> 00:19:26,600
Speaker 5: to create neutification images, this was the first time in

361
00:19:26,760 --> 00:19:31,520
Speaker 5: history that a newdifying tool was integrated into a widely

362
00:19:31,600 --> 00:19:36,000
Speaker 5: used social media platform, which combined the ability to just

363
00:19:36,480 --> 00:19:42,400
Speaker 5: generate an image with the widespread dissemination network, and so

364
00:19:43,080 --> 00:19:46,160
Speaker 5: an image could just be viewable by millions and millions

365
00:19:46,200 --> 00:19:49,600
Speaker 5: of people. But also, I mean, you know, x's for

366
00:19:49,720 --> 00:19:53,159
Speaker 5: people thirteen and up, and so children were also then

367
00:19:53,359 --> 00:19:58,040
Speaker 5: consumers of the images, which is illegal. And we're also

368
00:19:59,200 --> 00:20:01,119
Speaker 5: there were children that were the victims of it. And

369
00:20:01,359 --> 00:20:04,800
Speaker 5: the images posted were not just people in bikinis, but

370
00:20:05,080 --> 00:20:09,159
Speaker 5: GROC would pose them in sexual positions. It would respond

371
00:20:09,240 --> 00:20:12,879
Speaker 5: to prompts telling them to drench their body and blood

372
00:20:12,920 --> 00:20:17,920
Speaker 5: and seamen like fluids, whold suggestive props, and just really

373
00:20:18,040 --> 00:20:22,760
Speaker 5: vulgar and explicit images, and it seemed very clear that

374
00:20:22,800 --> 00:20:28,120
Speaker 5: this was an intended rollout a product that was known

375
00:20:28,160 --> 00:20:30,960
Speaker 5: to be harmful. They were like something like one point

376
00:20:30,960 --> 00:20:34,800
Speaker 5: eight million posts of women and children being deep faked,

377
00:20:35,160 --> 00:20:37,879
Speaker 5: which accounted for like forty one percent of all the

378
00:20:38,359 --> 00:20:42,480
Speaker 5: images posted during this nine day period. So Ashley Saint

379
00:20:42,480 --> 00:20:46,520
Speaker 5: Clair is somebody who is really demonized on the X

380
00:20:46,640 --> 00:20:49,440
Speaker 5: platform because of some of the hostile things that Elon

381
00:20:49,520 --> 00:20:52,119
Speaker 5: Musk has said about her, was a huge target of

382
00:20:52,160 --> 00:20:54,400
Speaker 5: the deep fakes, and there were just so many images

383
00:20:54,440 --> 00:20:59,720
Speaker 5: that GROC generated and published on her own page, and

384
00:20:59,760 --> 00:21:03,159
Speaker 5: so we sued them. We first got you attempted to

385
00:21:03,160 --> 00:21:07,640
Speaker 5: get a temporary restraining order to stop Grock from deepfaking her,

386
00:21:08,200 --> 00:21:11,600
Speaker 5: and then we sued them for defective design and other

387
00:21:11,960 --> 00:21:15,879
Speaker 5: sort of product liability types of lawsuits, negligence, and the

388
00:21:15,920 --> 00:21:17,920
Speaker 5: claim that I like the best is that we sue

389
00:21:17,920 --> 00:21:19,360
Speaker 5: them for public nuisance.

390
00:21:19,800 --> 00:21:22,359
Speaker 3: There's just no reasonable.

391
00:21:21,920 --> 00:21:26,080
Speaker 5: Safe use of this product but to harass people publicly.

392
00:21:26,200 --> 00:21:27,760
Speaker 3: So that's where it stands now.

393
00:21:27,840 --> 00:21:33,000
Speaker 5: And yeah, well, we had to give XAI statutory notice

394
00:21:33,000 --> 00:21:37,159
Speaker 5: that we were filing for a temporary restraining order, and

395
00:21:37,520 --> 00:21:42,280
Speaker 5: they immediately sued Ashley in Texas, claiming that she had

396
00:21:42,320 --> 00:21:46,560
Speaker 5: breached the XAI terms of service by even threatening to

397
00:21:46,600 --> 00:21:49,280
Speaker 5: sue them in her home state of New York. So

398
00:21:49,359 --> 00:21:53,120
Speaker 5: now we are litigating the issue in two different venues.

399
00:21:53,440 --> 00:21:57,280
Speaker 5: They're also trying to use the x terms of service

400
00:21:57,720 --> 00:22:01,800
Speaker 5: to get the XAI case transfer to New York, and

401
00:22:01,920 --> 00:22:05,440
Speaker 5: the judge initially last week said that XAI can rely

402
00:22:05,600 --> 00:22:08,880
Speaker 5: on the x terms of service to transfer.

403
00:22:08,480 --> 00:22:10,879
Speaker 3: The case to Texas. You mean to Texas.

404
00:22:10,960 --> 00:22:13,840
Speaker 5: Yeah, so even though x is not a defendant in

405
00:22:13,880 --> 00:22:17,800
Speaker 5: our case, only XAI is, the judge still let them

406
00:22:17,920 --> 00:22:21,879
Speaker 5: borrow from a different Elon company, which I think is

407
00:22:21,920 --> 00:22:22,600
Speaker 5: really scary.

408
00:22:22,800 --> 00:22:24,000
Speaker 3: So is that where it's going.

409
00:22:24,359 --> 00:22:27,840
Speaker 5: Well, we filed an emergency petition for Mandamus to try

410
00:22:27,840 --> 00:22:30,040
Speaker 5: to get the stay held up, but I don't know,

411
00:22:30,200 --> 00:22:31,400
Speaker 5: we don't have the outcome on that.

412
00:22:33,640 --> 00:22:45,360
Speaker 7: We'll be back in a minute.

413
00:22:47,960 --> 00:22:51,080
Speaker 3: So let's get to the tech infrastructure that supports the

414
00:22:51,119 --> 00:22:55,040
Speaker 3: creation dissemination of deep figs. Analysis from Media Outlet Indicator

415
00:22:55,280 --> 00:22:58,720
Speaker 3: found that newdify apps often use infrastructure from tech companies vs.

416
00:22:58,840 --> 00:23:02,920
Speaker 3: Explain the relationship between deep fake porn sites, tech companies,

417
00:23:02,920 --> 00:23:05,679
Speaker 3: and the social media platforms and to what extent do

418
00:23:05,760 --> 00:23:09,120
Speaker 3: these sites and newify apps rely on big tech infrastructure.

419
00:23:09,119 --> 00:23:11,840
Speaker 3: This was an issue when I was interviewing people who

420
00:23:11,880 --> 00:23:15,600
Speaker 3: were creating all manner of nonsense around the insurrection on

421
00:23:15,720 --> 00:23:19,240
Speaker 3: January sixth, and companies move quickly to take some of

422
00:23:19,280 --> 00:23:22,400
Speaker 3: it down, but they were clearly the way people got

423
00:23:22,440 --> 00:23:24,200
Speaker 3: to these different sites.

424
00:23:24,600 --> 00:23:27,679
Speaker 4: That's a broad question, Kara, so let me try and

425
00:23:27,720 --> 00:23:32,640
Speaker 4: break it down into pieces. The first is depending if

426
00:23:32,640 --> 00:23:37,160
Speaker 4: you're generating deep fakes at scale. So think the kind

427
00:23:37,160 --> 00:23:39,880
Speaker 4: of thing Laurie was talking about with mister deep fake.

428
00:23:40,240 --> 00:23:44,159
Speaker 4: He's trying to generate thousands, millions of deep fakes, and

429
00:23:44,240 --> 00:23:49,560
Speaker 4: so that requires computational hardware what are called graphical computing

430
00:23:49,680 --> 00:23:54,320
Speaker 4: units or GPUs, which today are pretty scarce and expensive.

431
00:23:55,040 --> 00:23:59,000
Speaker 4: So and that scarcity and price doesn't look like it's

432
00:23:59,080 --> 00:24:01,960
Speaker 4: coming down very much much anytime in the near future.

433
00:24:02,600 --> 00:24:09,200
Speaker 4: Those resources are heavily supported by cloud infrastructure providers. Again,

434
00:24:09,240 --> 00:24:12,119
Speaker 4: I'm not naming any specific companies, but there are a

435
00:24:12,200 --> 00:24:16,639
Speaker 4: number of companies that offer cloud services, so part of

436
00:24:16,680 --> 00:24:21,800
Speaker 4: the onus is placed on them if somebody is trying

437
00:24:21,840 --> 00:24:25,159
Speaker 4: to use their service to generate this kind of stuff

438
00:24:25,240 --> 00:24:28,080
Speaker 4: at scale. If somebody is trying to generate just a

439
00:24:28,080 --> 00:24:30,879
Speaker 4: handful of deep fakes, they're probably not going to figure

440
00:24:30,880 --> 00:24:34,200
Speaker 4: it out. But if somebody is generating millions and millions

441
00:24:34,240 --> 00:24:36,800
Speaker 4: of them, or thousands and thousands of them, they may

442
00:24:37,240 --> 00:24:39,919
Speaker 4: have some idea of what's going on. But that's not

443
00:24:40,080 --> 00:24:45,199
Speaker 4: easy either, because you know, there are many challenges for

444
00:24:45,320 --> 00:24:48,320
Speaker 4: some of these firms in figuring out what's a deep

445
00:24:48,320 --> 00:24:53,600
Speaker 4: fake and what's not, what's consensual, what's not and so

446
00:24:53,840 --> 00:24:57,200
Speaker 4: as the first issue, you know, the bad guys are

447
00:24:57,200 --> 00:25:01,280
Speaker 4: continuously evolving, so that me means that the actors in

448
00:25:01,359 --> 00:25:05,679
Speaker 4: question are changing. The technology used to generate this stuff

449
00:25:05,720 --> 00:25:10,200
Speaker 4: is changing, and it turns out, having studied deep fake

450
00:25:10,240 --> 00:25:13,960
Speaker 4: detectors quite a lot and created some on my own

451
00:25:14,520 --> 00:25:18,680
Speaker 4: with my lab, what we know is that small changes

452
00:25:19,400 --> 00:25:21,960
Speaker 4: can make a big difference. So to give you an example,

453
00:25:22,119 --> 00:25:25,159
Speaker 4: if I'm drinking this cup of coffee and there's a

454
00:25:25,200 --> 00:25:28,840
Speaker 4: deep fake of me with part of my face obscure

455
00:25:28,920 --> 00:25:32,359
Speaker 4: as I'm drinking that cup of coffee, it's harder in

456
00:25:32,400 --> 00:25:36,960
Speaker 4: many cases for many detectors to figure out that video

457
00:25:37,040 --> 00:25:40,000
Speaker 4: clip of me saying something which I probably never said

458
00:25:40,359 --> 00:25:45,400
Speaker 4: is not true. Likewise, the definition carry I'm not speaking

459
00:25:45,400 --> 00:25:49,679
Speaker 4: from a legal perspective, but for a normal person, what

460
00:25:49,920 --> 00:25:54,760
Speaker 4: constitutes an act that sexual in nature can vary a lot.

461
00:25:55,160 --> 00:25:58,879
Speaker 4: A sexualized act versus an actual act of you know,

462
00:25:59,040 --> 00:26:05,160
Speaker 4: some kind of sex very different. Also, the relative positions

463
00:26:05,200 --> 00:26:08,520
Speaker 4: in which a sexual act is carried out, where certain

464
00:26:08,560 --> 00:26:12,240
Speaker 4: body parts may be obscured but other things may be clear.

465
00:26:12,960 --> 00:26:16,480
Speaker 4: All that causes machine learning algorithms to detect deep fakes

466
00:26:16,720 --> 00:26:20,640
Speaker 4: to have a lot of problems. So I think those

467
00:26:20,680 --> 00:26:24,000
Speaker 4: are some of the challenges from a technical perspective in

468
00:26:24,080 --> 00:26:25,240
Speaker 4: figuring these things out.

469
00:26:25,400 --> 00:26:27,800
Speaker 3: So one analyst has said that some deep fake sites

470
00:26:27,800 --> 00:26:31,000
Speaker 3: now offer APIs to people creating non consensual images and

471
00:26:31,119 --> 00:26:33,639
Speaker 3: video generators explain what that means.

472
00:26:34,320 --> 00:26:38,280
Speaker 4: So, when you go to a website, you're typing stuff

473
00:26:38,320 --> 00:26:42,520
Speaker 4: in a prompt or a query, or clicking on some

474
00:26:42,560 --> 00:26:46,399
Speaker 4: buttons and you get a response. An API is an

475
00:26:46,440 --> 00:26:50,640
Speaker 4: application program interface, and it replaces you with a piece

476
00:26:50,680 --> 00:26:54,000
Speaker 4: of computer code, and that piece of computer code is

477
00:26:54,080 --> 00:26:57,919
Speaker 4: interacting in a different kind of language with the server

478
00:26:58,240 --> 00:27:02,680
Speaker 4: behind the website you're going to. And so if somebody

479
00:27:02,720 --> 00:27:07,200
Speaker 4: is using an API to generate deep fakes, that suggests

480
00:27:07,200 --> 00:27:10,080
Speaker 4: that they're generating it at scale. That suggests that they

481
00:27:10,119 --> 00:27:13,320
Speaker 4: have a piece of code that is shipping requests automatically

482
00:27:13,520 --> 00:27:16,239
Speaker 4: through that piece of code to the server, and that

483
00:27:16,320 --> 00:27:19,560
Speaker 4: server is sending back whatever they requested, in this case

484
00:27:19,960 --> 00:27:23,919
Speaker 4: some kind of sexualized image. That suggests somebody who's trying

485
00:27:23,920 --> 00:27:27,800
Speaker 4: to do it at scale, who is, you know, who

486
00:27:27,800 --> 00:27:30,800
Speaker 4: probably has his or her own databases of people they

487
00:27:30,840 --> 00:27:35,320
Speaker 4: want to propose in these situations that they will never

488
00:27:35,400 --> 00:27:36,560
Speaker 4: in to begin with.

489
00:27:37,400 --> 00:27:40,880
Speaker 3: Indicators Analysis also estimates that the new toifier economy may

490
00:27:40,920 --> 00:27:44,159
Speaker 3: be worth up to thirty six million dollars a year. Laurie,

491
00:27:44,200 --> 00:27:46,639
Speaker 3: and you're reporting on mister deep fakes. You noted the

492
00:27:46,640 --> 00:27:50,560
Speaker 3: financial incentives for the community members posting in the forum

493
00:27:50,720 --> 00:27:53,560
Speaker 3: tell us about the incentives and talk about the money

494
00:27:53,560 --> 00:27:56,520
Speaker 3: making Laurie here, and for the platforms that host this

495
00:27:56,680 --> 00:28:00,560
Speaker 3: content too, although for people don't realize. Goo Goal says

496
00:28:00,600 --> 00:28:03,680
Speaker 3: it doesn't allow apps to contain sexual content. Apple says

497
00:28:03,720 --> 00:28:07,400
Speaker 3: the company's apps to or prohibits overtly sexual content. Both

498
00:28:07,440 --> 00:28:11,400
Speaker 3: companies disabled searches for new defying their app stores. We'll

499
00:28:11,400 --> 00:28:13,399
Speaker 3: get to the workarounds in a second, but talk a

500
00:28:13,400 --> 00:28:16,040
Speaker 3: little bit aout the incentives, the financial incentives.

501
00:28:16,320 --> 00:28:19,040
Speaker 2: It's interesting because when we were looking at this, this

502
00:28:19,240 --> 00:28:21,320
Speaker 2: site that wasn't a site that was a platform. It

503
00:28:21,400 --> 00:28:24,440
Speaker 2: was a whole ecosystem that enabled the site to exist,

504
00:28:24,600 --> 00:28:28,080
Speaker 2: and that included platform liability from a lot of these

505
00:28:28,080 --> 00:28:31,439
Speaker 2: different companies. So for example, you know, the incentive was

506
00:28:31,480 --> 00:28:34,080
Speaker 2: you can go on not only can you view you know,

507
00:28:34,240 --> 00:28:37,119
Speaker 2: thousands of videos of sexually explicit deep fakes. If your

508
00:28:37,160 --> 00:28:39,680
Speaker 2: favorite celebrity, you can make your own. And this is

509
00:28:39,680 --> 00:28:42,400
Speaker 2: why we believed mister deep Fakes was so dangerous. He

510
00:28:42,600 --> 00:28:44,840
Speaker 2: was dangerous because he was creating an army of mister

511
00:28:44,880 --> 00:28:47,440
Speaker 2: deep fakes. Because now there are all these creators on there,

512
00:28:47,760 --> 00:28:49,600
Speaker 2: which is kind of pay to play, and there's a

513
00:28:49,640 --> 00:28:52,280
Speaker 2: freemium model, and then there's this premium model where you

514
00:28:52,320 --> 00:28:56,000
Speaker 2: can pay to have specialized deep fakes of public figures

515
00:28:56,000 --> 00:28:58,800
Speaker 2: who didn't consent and made and so people were probably

516
00:28:58,800 --> 00:29:01,240
Speaker 2: making more money than the guy behind mister deep fakes

517
00:29:01,240 --> 00:29:03,680
Speaker 2: off of this platform. It was creating a whole ecosystem.

518
00:29:03,680 --> 00:29:06,400
Speaker 2: But what else enabled that ecosystem? First, I would go

519
00:29:06,400 --> 00:29:09,840
Speaker 2: from a regulatory standpoint, the laws just hadn't caught up,

520
00:29:09,880 --> 00:29:13,320
Speaker 2: and we saw this with non consentual pornography. Oftentimes tech

521
00:29:13,360 --> 00:29:17,320
Speaker 2: moves quickly when there are the correct guardrails. Women and

522
00:29:17,520 --> 00:29:19,960
Speaker 2: young folks are the first ones to feel that pain.

523
00:29:20,360 --> 00:29:23,440
Speaker 2: But also there were payment providers that were accepting payments

524
00:29:23,520 --> 00:29:26,760
Speaker 2: on this platform, and so it was all of these things,

525
00:29:26,760 --> 00:29:30,000
Speaker 2: and we talk about these small steps that actually helped

526
00:29:30,120 --> 00:29:33,080
Speaker 2: enable this ecosystem to exist. And one of the coolest

527
00:29:33,120 --> 00:29:35,800
Speaker 2: things we saw because I think so many folks would say, well,

528
00:29:35,800 --> 00:29:37,440
Speaker 2: it's a game of whack a mole. You go after

529
00:29:37,480 --> 00:29:39,560
Speaker 2: this guy and then there are all these other ones

530
00:29:39,600 --> 00:29:41,960
Speaker 2: that are going to pop up, But actually what ended

531
00:29:42,040 --> 00:29:44,520
Speaker 2: up happening is we did go after this guy. We

532
00:29:44,600 --> 00:29:47,000
Speaker 2: found out who he was, and I think maybe most

533
00:29:47,000 --> 00:29:50,520
Speaker 2: alarming was he wasn't a technical guy. This is a

534
00:29:50,560 --> 00:29:54,120
Speaker 2: guy who was, i would say, in his thirties, worked

535
00:29:54,160 --> 00:29:57,800
Speaker 2: at a pharmacist at a hospital, helping people, was newly married,

536
00:29:57,960 --> 00:30:01,040
Speaker 2: had a young child, and he had been able to

537
00:30:01,120 --> 00:30:05,400
Speaker 2: create a website that enabled a whole new wave.

538
00:30:05,200 --> 00:30:07,120
Speaker 1: Of deep fake abuse because.

539
00:30:06,840 --> 00:30:09,960
Speaker 2: He had an interest in technology and deep fakes, and

540
00:30:10,000 --> 00:30:13,240
Speaker 2: he saw an opening because when Reddit banned sexually explicit

541
00:30:13,320 --> 00:30:16,480
Speaker 2: deep fakes, we were able to track where his username

542
00:30:16,840 --> 00:30:20,360
Speaker 2: was talking about seeing an opportunity that could be profitable

543
00:30:20,640 --> 00:30:23,440
Speaker 2: and getting other people to build on this. And so

544
00:30:23,720 --> 00:30:26,880
Speaker 2: all of these things together created i would say, an

545
00:30:26,880 --> 00:30:30,480
Speaker 2: ecosystem that allowed it. And all of these things also

546
00:30:31,080 --> 00:30:33,800
Speaker 2: once people started chipping away at this, once the laws

547
00:30:33,840 --> 00:30:38,480
Speaker 2: started slowly but surely catching up, once Google deranked the platform,

548
00:30:38,600 --> 00:30:41,600
Speaker 2: there were all sorts of small steps that actually created

549
00:30:41,600 --> 00:30:43,760
Speaker 2: a friction that made it harder to exist.

550
00:30:44,000 --> 00:30:46,520
Speaker 3: So let's shift to the current laws and regulations surrounding

551
00:30:46,520 --> 00:30:48,760
Speaker 3: deep fake porn, including the Take It Down Act, the

552
00:30:48,840 --> 00:30:51,000
Speaker 3: law that makes it a federal crime to publish non

553
00:30:51,000 --> 00:30:54,640
Speaker 3: consensual explicit deep fakes. The federal government began enforcing it

554
00:30:54,720 --> 00:30:58,040
Speaker 3: in May, Carrie, in your pursuit of tech platform accountability.

555
00:30:58,040 --> 00:31:00,640
Speaker 3: You've cut up against section two thirty. Have many the

556
00:31:00,720 --> 00:31:03,800
Speaker 3: law that protects platforms for being held legally liable for

557
00:31:03,880 --> 00:31:07,680
Speaker 3: content that users post. You mentioned product liability law as

558
00:31:07,720 --> 00:31:10,800
Speaker 3: a way to stop deep fake porn and get justice

559
00:31:10,800 --> 00:31:14,400
Speaker 3: for victims. So talk about using product liability law.

560
00:31:14,720 --> 00:31:19,480
Speaker 5: Section two thirty was intended to basically just immunize platforms

561
00:31:19,520 --> 00:31:23,600
Speaker 5: for content that third parties created and posted. But when

562
00:31:23,600 --> 00:31:27,920
Speaker 5: it comes to deep fakes, especially like in the situation

563
00:31:28,000 --> 00:31:31,160
Speaker 5: of Groc, we're saying that Grock was the one who

564
00:31:31,560 --> 00:31:35,520
Speaker 5: created this material and also published it, and therefore section

565
00:31:35,560 --> 00:31:39,320
Speaker 5: two thirty shouldn't apply at all. But even if it does,

566
00:31:40,040 --> 00:31:43,600
Speaker 5: by suing them under product liability, we're suing them not

567
00:31:44,080 --> 00:31:49,920
Speaker 5: for publishing the content, but for you know, basically harming

568
00:31:49,960 --> 00:31:54,360
Speaker 5: people through its defective design, defective manufacturing.

569
00:31:54,520 --> 00:31:56,640
Speaker 3: These were all foreseeable uses.

570
00:31:56,680 --> 00:32:01,960
Speaker 5: There's no reasonably safe use for new toifying product, and

571
00:32:02,000 --> 00:32:05,200
Speaker 5: so that's just like a second way to overcome section

572
00:32:05,240 --> 00:32:11,480
Speaker 5: two thirty. Now, XAI, they haven't fully exposed what their

573
00:32:11,520 --> 00:32:13,800
Speaker 5: defenses are going to be in our case, but when

574
00:32:13,840 --> 00:32:17,760
Speaker 5: they were opposing the temporary restraining order, they were claiming

575
00:32:17,840 --> 00:32:20,560
Speaker 5: that they wanted. I mean, so far, this would be

576
00:32:20,640 --> 00:32:23,520
Speaker 5: the only AI product that is saying that Section two

577
00:32:23,560 --> 00:32:27,560
Speaker 5: thirty should apply to it and that they're not responsible

578
00:32:27,880 --> 00:32:32,880
Speaker 5: for what people type in as prompts. But even more

579
00:32:32,880 --> 00:32:35,880
Speaker 5: alarming is that they're claiming that they should have the

580
00:32:35,920 --> 00:32:40,520
Speaker 5: First Amendment protections and that rock should basically be protected

581
00:32:40,920 --> 00:32:45,160
Speaker 5: under free speech, which is a pretty alarming idea. You

582
00:32:45,160 --> 00:32:50,680
Speaker 5: know that AI and a chatbot would have free speech rights.

583
00:32:51,560 --> 00:32:51,960
Speaker 3: They don't.

584
00:32:52,040 --> 00:32:54,000
Speaker 5: You know, so far, we're not at a point where

585
00:32:54,000 --> 00:32:58,200
Speaker 5: we're giving constitutional rights to a chatbot. But again, the

586
00:32:58,360 --> 00:33:01,320
Speaker 5: arguments haven't yet hit this the stage where there's been

587
00:33:01,360 --> 00:33:02,160
Speaker 5: fully brief yet.

588
00:33:02,680 --> 00:33:05,240
Speaker 3: So vas talk about workarounds because a lot of these

589
00:33:05,840 --> 00:33:08,920
Speaker 3: these companies are very clever in terms of workarounds. Porn

590
00:33:08,920 --> 00:33:11,600
Speaker 3: people have always been very clever. How easy is it

591
00:33:11,640 --> 00:33:14,480
Speaker 3: to do workarounds here from a technological point of view?

592
00:33:15,400 --> 00:33:18,160
Speaker 4: You know, I was thinking about this when Lori was

593
00:33:18,200 --> 00:33:22,040
Speaker 4: speaking earlier Kara. You know, she talked about a website

594
00:33:22,320 --> 00:33:25,000
Speaker 4: where somebody goes in and says, I want to create

595
00:33:25,040 --> 00:33:28,600
Speaker 4: a deep fake of such and such person doing such

596
00:33:28,640 --> 00:33:33,080
Speaker 4: and such thing. And so that's a classic example off

597
00:33:33,640 --> 00:33:38,320
Speaker 4: a place where that website is using the APIs that

598
00:33:38,400 --> 00:33:43,800
Speaker 4: you asked about earlier to access a cloud provider where

599
00:33:43,920 --> 00:33:48,720
Speaker 4: this code running degenerate the newtification of the sort designed

600
00:33:48,720 --> 00:33:52,560
Speaker 4: by whoever asked the query. But now think about it

601
00:33:52,600 --> 00:33:57,840
Speaker 4: this way. Let's say the platform has the mechanism. Let's

602
00:33:57,880 --> 00:34:00,440
Speaker 4: say they have a perfect mechanism that doesn't exis yes today,

603
00:34:00,440 --> 00:34:03,520
Speaker 4: but suppose they did to detect the deep fake video

604
00:34:03,720 --> 00:34:08,400
Speaker 4: of some form of sexual activity. Then the workarounds for

605
00:34:08,480 --> 00:34:11,200
Speaker 4: the porn guys that are some simple ones. So one

606
00:34:11,239 --> 00:34:14,880
Speaker 4: of the classic methods to generate deep fakes, regardless of

607
00:34:14,880 --> 00:34:18,840
Speaker 4: whether it's for good or bad, is something called stable diffusion.

608
00:34:19,560 --> 00:34:24,000
Speaker 4: So in stable diffusion, what you do is you, let's say,

609
00:34:24,000 --> 00:34:26,000
Speaker 4: throw in an image. I'll use an image rather than

610
00:34:26,000 --> 00:34:28,480
Speaker 4: a video as my example, but you throw in an

611
00:34:28,480 --> 00:34:32,759
Speaker 4: image of somebody and that image gets converted to some

612
00:34:33,000 --> 00:34:36,720
Speaker 4: very coarse representation. You know, think of this as something

613
00:34:36,760 --> 00:34:40,640
Speaker 4: that's a very skeletal version of that image, and then

614
00:34:41,040 --> 00:34:46,080
Speaker 4: what ends up happening is that skeletal version goes through

615
00:34:46,080 --> 00:34:51,280
Speaker 4: the neuification process, but at the skeletal level, so that's

616
00:34:51,800 --> 00:34:58,399
Speaker 4: skeletalized neutified image doesn't look anything like what a real

617
00:34:58,560 --> 00:35:01,759
Speaker 4: nude might look like, and so you can think of

618
00:35:01,800 --> 00:35:07,799
Speaker 4: it as some kind of computationally weird representation of, you know,

619
00:35:07,880 --> 00:35:11,440
Speaker 4: a person who's about to be sexually abused via a

620
00:35:11,480 --> 00:35:14,880
Speaker 4: deep fake. And then they could do all this on platform,

621
00:35:15,040 --> 00:35:18,520
Speaker 4: on a cloud, and then without ever passing the source

622
00:35:18,520 --> 00:35:22,920
Speaker 4: image or the final image, they bring back this course

623
00:35:23,000 --> 00:35:26,759
Speaker 4: representation of the new deified image and perform the last

624
00:35:26,800 --> 00:35:29,600
Speaker 4: step back on their own servers, where it generates the

625
00:35:29,600 --> 00:35:34,080
Speaker 4: final nude. So that's an example of one way in

626
00:35:34,120 --> 00:35:37,680
Speaker 4: which somebody with you know, just limited computer science knowledge,

627
00:35:38,040 --> 00:35:41,520
Speaker 4: who's familiar with the code, who has access to all

628
00:35:41,600 --> 00:35:45,920
Speaker 4: this code, can generate something while avoiding the guardrails of

629
00:35:46,000 --> 00:35:48,920
Speaker 4: some of the platforms, even if those guardrails are close

630
00:35:49,000 --> 00:35:49,560
Speaker 4: to perfect.

631
00:35:51,360 --> 00:36:06,799
Speaker 3: We'll be back in a minute. The Defiance Act, which

632
00:36:06,840 --> 00:36:09,560
Speaker 3: is aimed at allowing victims of non consensual deep figs

633
00:36:09,560 --> 00:36:11,719
Speaker 3: to suit for damages, passed in the Senate, and it's

634
00:36:11,760 --> 00:36:14,880
Speaker 3: currently stalled in the House. Even though some states have

635
00:36:14,960 --> 00:36:19,839
Speaker 3: criminalized deep figs, enforcement remains low. Laurie, What extent are

636
00:36:19,840 --> 00:36:22,640
Speaker 3: these laws effective when it comes to stopping or at

637
00:36:22,680 --> 00:36:26,120
Speaker 3: least curbing the spread. It's happened before with credit card

638
00:36:26,120 --> 00:36:29,960
Speaker 3: companies and certain businesses. You know, things tend to slow

639
00:36:30,040 --> 00:36:32,799
Speaker 3: things down. But in this case, how important is it

640
00:36:32,800 --> 00:36:35,240
Speaker 3: to have the Take It Down and the Defiance Act?

641
00:36:35,760 --> 00:36:38,040
Speaker 2: I'll answer, and I'm curious for Carry's response too, but

642
00:36:38,040 --> 00:36:40,840
Speaker 2: I think it's incredibly important. You know, the Take It

643
00:36:40,920 --> 00:36:44,120
Speaker 2: Down Act, you know, force tech companies to pay attention

644
00:36:44,200 --> 00:36:46,960
Speaker 2: and quicker. It gives them forty eight hours to take

645
00:36:47,000 --> 00:36:51,480
Speaker 2: down non consensual abuse. But oftentimes when it comes to victims,

646
00:36:51,760 --> 00:36:54,719
Speaker 2: it's really difficult to find recourse for this, you know,

647
00:36:54,800 --> 00:36:57,840
Speaker 2: And so the Defiance Act would create civil penalties that

648
00:36:57,840 --> 00:37:01,040
Speaker 2: would enable them to actually have the ability to go

649
00:37:01,120 --> 00:37:03,680
Speaker 2: after the folks who have done this and have civil recourse,

650
00:37:03,719 --> 00:37:06,040
Speaker 2: which I think makes a big difference. And when we

651
00:37:06,160 --> 00:37:08,440
Speaker 2: started out on our search for mister Deep Fakes, which was,

652
00:37:08,560 --> 00:37:11,960
Speaker 2: let's just be honest, really a search for asking why

653
00:37:12,040 --> 00:37:14,759
Speaker 2: is this allowed to happen? You know, there weren't a

654
00:37:14,760 --> 00:37:16,440
Speaker 2: lot of state laws, and I will say a lot

655
00:37:16,440 --> 00:37:18,240
Speaker 2: of the state laws and a lot of the victims

656
00:37:18,239 --> 00:37:22,120
Speaker 2: and survivors we spoke to. Unfortunately, the onus in many

657
00:37:22,160 --> 00:37:25,120
Speaker 2: ways becomes on these survivors to speak out about it

658
00:37:25,160 --> 00:37:28,000
Speaker 2: and help change laws, and they did in the state level,

659
00:37:28,080 --> 00:37:31,080
Speaker 2: which you know, the state laws changing all of them,

660
00:37:31,239 --> 00:37:34,960
Speaker 2: varying in scope. Civil criminal penalties actually were where a

661
00:37:35,000 --> 00:37:38,080
Speaker 2: lot of these different victims initially had recourse. Which is

662
00:37:38,080 --> 00:37:40,960
Speaker 2: also scary to watch as folks go after regulation at

663
00:37:40,960 --> 00:37:43,680
Speaker 2: the state level because oftentimes the federal government is just

664
00:37:43,719 --> 00:37:45,920
Speaker 2: slow to move. I would say, I hope that the

665
00:37:45,920 --> 00:37:49,239
Speaker 2: Defiance Act passes, it can get through. I think this

666
00:37:49,280 --> 00:37:51,960
Speaker 2: is a bipartisan issue because it's really a child safety

667
00:37:52,000 --> 00:37:53,000
Speaker 2: issue as well.

668
00:37:53,120 --> 00:37:54,719
Speaker 3: Right, which is how they're saying it. And so, by

669
00:37:54,719 --> 00:37:57,160
Speaker 3: the way, Wired investigation found that GROC is still being

670
00:37:57,239 --> 00:38:01,560
Speaker 3: used for non consensual explicit content, this despite regulatory scrutiny

671
00:38:01,560 --> 00:38:05,160
Speaker 3: and multiple lawsuits, and after Musk's XAI said it would

672
00:38:05,200 --> 00:38:08,839
Speaker 3: add restrictions in response to an NBC investigation, the x

673
00:38:08,880 --> 00:38:12,000
Speaker 3: safety account post, do we strictly prohibit users from generating

674
00:38:12,080 --> 00:38:15,240
Speaker 3: non consensual explicit deep fakes and from using our tools

675
00:38:15,239 --> 00:38:19,680
Speaker 3: to undress real people? Explain how some of these legal

676
00:38:19,760 --> 00:38:23,000
Speaker 3: dynamics apply to XAI in that regard. Since you're in

677
00:38:23,040 --> 00:38:25,800
Speaker 3: a lawsuit with them, when it keeps showing up that

678
00:38:25,840 --> 00:38:29,240
Speaker 3: they're doing the same thing, you would accuse them of Well.

679
00:38:29,040 --> 00:38:31,520
Speaker 5: I mean that statement that they say that it's against

680
00:38:31,600 --> 00:38:35,640
Speaker 5: their regulations and they're basically passing the blame to users.

681
00:38:36,080 --> 00:38:39,120
Speaker 5: There's nothing in that statement that suggests that they are

682
00:38:39,800 --> 00:38:44,040
Speaker 5: modifying their product so that deep fakes are no longer

683
00:38:44,520 --> 00:38:47,560
Speaker 5: easy to create. And the problem with things like to

684
00:38:47,640 --> 00:38:50,080
Speaker 5: take it down act it's great, you know, suddenly now

685
00:38:50,239 --> 00:38:52,759
Speaker 5: there's a forty eight hour window of time that tech

686
00:38:52,760 --> 00:38:57,400
Speaker 5: platforms are supposed to remove illegal content, But it has

687
00:38:57,440 --> 00:39:02,440
Speaker 5: no teeth because victims can't enforce it. Victims can't soothe

688
00:39:02,520 --> 00:39:05,440
Speaker 5: the platform and say, hey, you know, it's now been

689
00:39:05,560 --> 00:39:08,160
Speaker 5: forty eight days and that content is still up. So

690
00:39:08,360 --> 00:39:11,520
Speaker 5: the problem with all regulations and all these options is

691
00:39:11,520 --> 00:39:15,359
Speaker 5: that the victims cannot enforce them. And so when we're

692
00:39:15,400 --> 00:39:18,160
Speaker 5: talking about who's the bad guy? VS has talked about

693
00:39:18,200 --> 00:39:21,120
Speaker 5: the bad guys in these situations, in my view, the

694
00:39:21,160 --> 00:39:25,520
Speaker 5: worst guys are the platforms. And so until we have

695
00:39:25,680 --> 00:39:29,000
Speaker 5: regulations that put the victims in the driver's seats to

696
00:39:29,040 --> 00:39:34,080
Speaker 5: not just you know, get criminal justice against the individuals,

697
00:39:34,160 --> 00:39:38,400
Speaker 5: but to actually get justice against the platforms, you know,

698
00:39:38,480 --> 00:39:41,640
Speaker 5: I think the only recourse we have is to use

699
00:39:41,680 --> 00:39:43,879
Speaker 5: our courts and make these companies pay.

700
00:39:44,360 --> 00:39:46,880
Speaker 3: So every episode we get a question from an outside expert.

701
00:39:46,920 --> 00:39:49,160
Speaker 3: Here's yours, and you're all going to answer it, So

702
00:39:49,280 --> 00:39:51,439
Speaker 3: let's listen to it first. Hi.

703
00:39:51,719 --> 00:39:54,320
Speaker 8: I'm Georgia Wells, a tech reporter for The Wall Street Journal.

704
00:39:54,800 --> 00:39:56,799
Speaker 8: I just finished a deep dive into the issue of

705
00:39:56,880 --> 00:40:00,080
Speaker 8: teens using deep fake nudes to harass each other, and

706
00:40:00,120 --> 00:40:02,160
Speaker 8: so the big question I would ask is that, now

707
00:40:02,160 --> 00:40:04,640
Speaker 8: that the government has just started enforcing the Take It

708
00:40:04,680 --> 00:40:08,279
Speaker 8: Down Act, how might this change the harassment landscape for

709
00:40:08,360 --> 00:40:09,200
Speaker 8: young people today?

710
00:40:09,960 --> 00:40:14,640
Speaker 3: Thank you, Laura, you go first, then Carrie then VS. Yeah.

711
00:40:14,719 --> 00:40:17,000
Speaker 2: I do agree a little bit with Carrie on saying,

712
00:40:17,120 --> 00:40:19,399
Speaker 2: you know, there's only so much you can actually kind

713
00:40:19,400 --> 00:40:22,160
Speaker 2: of enforce when the onus is on the victim. I

714
00:40:22,239 --> 00:40:26,480
Speaker 2: do think from a cultural standpoint, it is very different

715
00:40:26,480 --> 00:40:28,480
Speaker 2: than it was a couple of years ago of federal

716
00:40:28,520 --> 00:40:31,640
Speaker 2: law Take It Down Act. Again, state laws are at

717
00:40:31,719 --> 00:40:35,480
Speaker 2: least sending this message that the stakes will be high

718
00:40:35,520 --> 00:40:39,439
Speaker 2: and that young boys, young teenagers can get into real

719
00:40:39,480 --> 00:40:41,719
Speaker 2: trouble for doing this and can get charged. I mean,

720
00:40:42,320 --> 00:40:45,680
Speaker 2: I think that at least from a perspective of schools

721
00:40:45,719 --> 00:40:48,160
Speaker 2: being able to speak to students, parents being able to

722
00:40:48,200 --> 00:40:51,200
Speaker 2: speak to their children about this, and having some structure

723
00:40:51,239 --> 00:40:54,479
Speaker 2: will be helpful again, though I'm not sure the Take It.

724
00:40:54,400 --> 00:40:56,160
Speaker 1: Down Act is going to be the thing that's going

725
00:40:56,200 --> 00:40:57,040
Speaker 1: to actually change this.

726
00:40:57,160 --> 00:40:59,759
Speaker 2: I also agree with Kerrie on platform liability in creating

727
00:40:59,800 --> 00:41:03,959
Speaker 2: high stakes for the tech companies and the whole ecosystem

728
00:41:03,960 --> 00:41:05,520
Speaker 2: that enables this to happen.

729
00:41:05,520 --> 00:41:08,920
Speaker 5: Kerrie, So I think they'll get the memo that this

730
00:41:09,080 --> 00:41:13,319
Speaker 5: is illegal. I mean, having criminal laws was really really

731
00:41:13,360 --> 00:41:17,200
Speaker 5: effective when we were dealing with the scourge of image

732
00:41:17,239 --> 00:41:20,520
Speaker 5: asexual abuse. Over the last decade. We went from having

733
00:41:20,520 --> 00:41:23,680
Speaker 5: three states to having fifty states with those laws, and

734
00:41:23,920 --> 00:41:27,879
Speaker 5: my firm saw such an enormous and quick plummet of

735
00:41:27,880 --> 00:41:32,839
Speaker 5: offenders creating images like that. So that's great, But we

736
00:41:32,880 --> 00:41:37,520
Speaker 5: still don't have an apparatus that targets the platforms themselves

737
00:41:37,840 --> 00:41:43,279
Speaker 5: and prevents them from basically monetizing this kind of humiliation.

738
00:41:43,960 --> 00:41:48,120
Speaker 5: And until we have laws that victims can enforce themselves, like,

739
00:41:48,280 --> 00:41:53,360
Speaker 5: it's not really harming they who I consider the true offenders.

740
00:41:53,000 --> 00:41:56,120
Speaker 3: Right, which are the platforms themselves, all right? Vs.

741
00:41:56,280 --> 00:42:01,360
Speaker 4: You know, I have very mixed feelings about criminalizing a

742
00:42:01,440 --> 00:42:04,600
Speaker 4: thirteen year old, all right, But at the same time,

743
00:42:05,080 --> 00:42:08,080
Speaker 4: we need to protect perhaps other thirteen year olds or

744
00:42:08,160 --> 00:42:11,560
Speaker 4: fourteen year olds or ten year olds from the consequences

745
00:42:12,080 --> 00:42:14,640
Speaker 4: that accrue to them because of some thirteen year old

746
00:42:15,120 --> 00:42:18,879
Speaker 4: putting out deep porn of that kid. So I think,

747
00:42:19,000 --> 00:42:22,160
Speaker 4: you know, a first step has got to be something

748
00:42:22,800 --> 00:42:28,279
Speaker 4: around education. A lot of our teachers in schools are

749
00:42:28,600 --> 00:42:34,839
Speaker 4: simply not teaching their kids about deep fakes. And it's

750
00:42:34,920 --> 00:42:38,280
Speaker 4: one thing if kids are using deep fakes or generative

751
00:42:38,320 --> 00:42:43,040
Speaker 4: AI to you know, create reports for their class projects.

752
00:42:43,040 --> 00:42:44,680
Speaker 4: I mean that brings up a whole bunch of other

753
00:42:44,680 --> 00:42:48,560
Speaker 4: ethical issues. But it's another thing if they use deep

754
00:42:48,600 --> 00:42:54,680
Speaker 4: fakes to explicitly cause damage to a kid and one

755
00:42:54,719 --> 00:42:58,040
Speaker 4: of their classmates. So I think the ethics around us

756
00:42:58,400 --> 00:43:01,600
Speaker 4: has got to be clearly or take in class and

757
00:43:01,840 --> 00:43:04,520
Speaker 4: by schools. And some of us schools may have done

758
00:43:04,520 --> 00:43:08,080
Speaker 4: a good job of this, others may not. And you

759
00:43:08,080 --> 00:43:13,239
Speaker 4: know that education and the consequences of those actions in

760
00:43:13,280 --> 00:43:17,280
Speaker 4: a school context are clear. I think it'll be challenging.

761
00:43:17,560 --> 00:43:20,120
Speaker 2: You know, viaz just to say to respond if it's helpful,

762
00:43:20,160 --> 00:43:22,920
Speaker 2: just to respond to that. I think the frustration is

763
00:43:23,000 --> 00:43:25,920
Speaker 2: so many times, I agree, how do we expect thirteen

764
00:43:26,000 --> 00:43:29,080
Speaker 2: year old boys to know that this isn't okay when

765
00:43:29,080 --> 00:43:32,000
Speaker 2: they're getting served up neuification apps on social media.

766
00:43:32,480 --> 00:43:33,799
Speaker 1: You know, I think it's it is.

767
00:43:34,120 --> 00:43:36,360
Speaker 2: It's an end like two things can be true, and

768
00:43:36,440 --> 00:43:39,359
Speaker 2: so there's a huge gap between education and being able

769
00:43:39,440 --> 00:43:42,440
Speaker 2: to educate folks on this and what the platforms are

770
00:43:42,480 --> 00:43:45,240
Speaker 2: allowing and what they say they're allowing and what's coming through.

771
00:43:45,360 --> 00:43:47,600
Speaker 1: It creates a really difficult environment.

772
00:43:48,000 --> 00:43:52,400
Speaker 4: You know, I worry generally about the broad statement or

773
00:43:52,480 --> 00:43:57,760
Speaker 4: the broad platforms Moniker. I think there's a wide variety

774
00:43:58,200 --> 00:44:01,680
Speaker 4: in what different platforms are doing. Some platforms are going

775
00:44:01,719 --> 00:44:05,040
Speaker 4: to a great amount of trouble, spending a tremendous amount

776
00:44:05,080 --> 00:44:09,360
Speaker 4: of money to try and perform the best innovation possible

777
00:44:09,760 --> 00:44:12,640
Speaker 4: so they can put some of these problems to bed,

778
00:44:13,160 --> 00:44:16,040
Speaker 4: and others are doing much much less. So there's like

779
00:44:16,080 --> 00:44:20,000
Speaker 4: a huge range of what platforms are doing, some doing

780
00:44:20,040 --> 00:44:23,719
Speaker 4: relatively little, some making every effort to try and get there.

781
00:44:24,280 --> 00:44:28,719
Speaker 4: Now I want to add that, you know, again, even

782
00:44:28,760 --> 00:44:31,840
Speaker 4: if platforms who are doing their very best, there's a

783
00:44:31,840 --> 00:44:35,920
Speaker 4: famous saying, due to a statistician at the University of Wisconsin,

784
00:44:36,719 --> 00:44:40,319
Speaker 4: all models are wrong, some are useful, all right, And

785
00:44:40,719 --> 00:44:44,680
Speaker 4: so whatever models are being used by platforms to find

786
00:44:44,719 --> 00:44:48,160
Speaker 4: these deep fakes. They're never going to be perfect, and

787
00:44:48,239 --> 00:44:51,399
Speaker 4: so the question is are they making the effort? Some

788
00:44:51,440 --> 00:44:54,319
Speaker 4: are making what I would consider a very good faith effort.

789
00:44:54,800 --> 00:44:57,919
Speaker 3: Some who is doing a good job.

790
00:44:58,000 --> 00:45:00,200
Speaker 4: I'd prefer to stay out of that, but I don't

791
00:45:00,200 --> 00:45:01,839
Speaker 4: want to talk about specific platforms.

792
00:45:01,880 --> 00:45:04,960
Speaker 3: But yeah, so let's end by looking at what individuals

793
00:45:04,960 --> 00:45:08,800
Speaker 3: can do as deep fake technology continues to evolve. Carrying

794
00:45:08,840 --> 00:45:11,200
Speaker 3: now that the Take It Down Act forces tech companies

795
00:45:11,239 --> 00:45:13,680
Speaker 3: to remove deep fake nudes, how do you request a

796
00:45:13,800 --> 00:45:16,040
Speaker 3: takedown if you're targeted? Well, part of what.

797
00:45:15,960 --> 00:45:20,520
Speaker 5: The Takedown Act required was that there be a flow

798
00:45:21,239 --> 00:45:26,080
Speaker 5: for people to easily request content removal, and so that

799
00:45:26,200 --> 00:45:29,839
Speaker 5: kind of infrastructure I really appreciated that being incorporated into

800
00:45:29,880 --> 00:45:33,520
Speaker 5: the laws. So ideally, you know, a victim within the

801
00:45:33,880 --> 00:45:37,120
Speaker 5: platform itself can now report it, you know, whether it's

802
00:45:37,120 --> 00:45:41,080
Speaker 5: through a URL or if it's like a Instagram or Facebook,

803
00:45:41,080 --> 00:45:44,600
Speaker 5: then the image itself would ideally have a method to

804
00:45:44,800 --> 00:45:46,600
Speaker 5: request removal.

805
00:45:47,040 --> 00:45:50,680
Speaker 3: When digital forensics professor and deep fake expert Hani Farred

806
00:45:50,840 --> 00:45:53,560
Speaker 3: was on the podcast in January, he expressed concerns that

807
00:45:53,680 --> 00:45:56,920
Speaker 3: agentic AI will supercharge the creation of deep figs in

808
00:45:56,960 --> 00:45:59,400
Speaker 3: the coming months vs. How do we make sure the

809
00:45:59,440 --> 00:46:02,840
Speaker 3: new laws policies don't block positive uses in these tools.

810
00:46:03,320 --> 00:46:05,160
Speaker 4: So I think, again, if we don't want to block

811
00:46:05,400 --> 00:46:09,400
Speaker 4: the technology, we want the technology to evolve. Because again,

812
00:46:10,160 --> 00:46:14,000
Speaker 4: as Laurie said much earlier in the podcast today, you know,

813
00:46:14,560 --> 00:46:18,920
Speaker 4: deep fakes do have the capacity to help do things

814
00:46:19,040 --> 00:46:22,080
Speaker 4: like better understand disease. I'm not going to call this

815
00:46:22,160 --> 00:46:25,680
Speaker 4: a deep fake, but the generative AI is being used

816
00:46:25,680 --> 00:46:29,239
Speaker 4: to create new kinds of proteins that will help cure diseases.

817
00:46:29,600 --> 00:46:33,040
Speaker 4: Generative AI is being used to create materials that will

818
00:46:33,080 --> 00:46:35,799
Speaker 4: degrade and perform in ways that we would like, that

819
00:46:35,840 --> 00:46:38,520
Speaker 4: would protect the environment. And more so, I think what

820
00:46:38,560 --> 00:46:41,759
Speaker 4: we have to do is to look at intent. You know,

821
00:46:42,239 --> 00:46:46,160
Speaker 4: what is the intent of the individual who is creating

822
00:46:46,200 --> 00:46:49,960
Speaker 4: a deep fake. And in the case of the thirteen

823
00:46:50,040 --> 00:46:52,800
Speaker 4: year old boy or fourteen year old boy who's creating

824
00:46:53,000 --> 00:46:56,880
Speaker 4: deep fakes of one of his classmates because he's pissed

825
00:46:56,920 --> 00:47:03,080
Speaker 4: off about something, I think they're there's intent. Whereas if

826
00:47:03,080 --> 00:47:05,880
Speaker 4: somebody you know, there was this guy who created a

827
00:47:05,920 --> 00:47:09,360
Speaker 4: deep fake of the Pope but Francis in puffy robes

828
00:47:09,400 --> 00:47:12,839
Speaker 4: and a very dapper looking version of him. You know

829
00:47:13,160 --> 00:47:16,440
Speaker 4: that I think is something that's laughable, and I find

830
00:47:16,480 --> 00:47:19,719
Speaker 4: that at least I find that somewhat entertaining. I doubt

831
00:47:19,800 --> 00:47:22,880
Speaker 4: very much of the Pope was, you know, horribly offended.

832
00:47:23,400 --> 00:47:26,000
Speaker 4: And so, you know, I think we want to allow

833
00:47:26,120 --> 00:47:31,319
Speaker 4: artistic freedom and expression scientific inquiry at the same time

834
00:47:31,560 --> 00:47:35,279
Speaker 4: regulate so that it's the uses that are banned or

835
00:47:35,840 --> 00:47:36,520
Speaker 4: that lead.

836
00:47:36,360 --> 00:47:39,759
Speaker 3: To the individual intent one. Yeah, Laurie, In the end,

837
00:47:39,800 --> 00:47:42,799
Speaker 3: you track down mister deep fakes, what ultimately brought down

838
00:47:42,800 --> 00:47:44,200
Speaker 3: the site? And where is he now?

839
00:47:44,600 --> 00:47:47,400
Speaker 1: Great question? Where is he now? I would love to know.

840
00:47:47,640 --> 00:47:50,920
Speaker 2: So eventually we did track him down with the help

841
00:47:51,040 --> 00:47:54,840
Speaker 2: of you know, investigative journalists the internet and some of

842
00:47:54,840 --> 00:47:57,600
Speaker 2: our cybersecurity friends who did some pro bono work for

843
00:47:57,680 --> 00:47:59,799
Speaker 2: us because they have kids and they viewed this as

844
00:47:59,800 --> 00:48:02,439
Speaker 2: a threat. And we were able to track him down,

845
00:48:02,719 --> 00:48:05,040
Speaker 2: and you know, Carol, like I'll never forget confronting him

846
00:48:05,120 --> 00:48:07,720
Speaker 2: on his way into the office. And I do remember

847
00:48:07,760 --> 00:48:11,040
Speaker 2: a Carrie who I've known because I've covered nonconsensual pornography before.

848
00:48:11,080 --> 00:48:13,759
Speaker 2: I called her before going to his parents' home to

849
00:48:13,800 --> 00:48:16,480
Speaker 2: make sure that we did everything correct, just in case,

850
00:48:16,719 --> 00:48:19,560
Speaker 2: you know, we didn't want to create any legal liabilities.

851
00:48:19,920 --> 00:48:22,360
Speaker 2: But I will never forget him looking at me. He

852
00:48:22,440 --> 00:48:25,359
Speaker 2: knew who I was with so much disdain. And I

853
00:48:25,400 --> 00:48:27,759
Speaker 2: know this is a delusional part of me, but there

854
00:48:27,840 --> 00:48:29,680
Speaker 2: was a part of me that thought maybe we would

855
00:48:29,719 --> 00:48:32,000
Speaker 2: have tracked this guy down and we could show up

856
00:48:32,040 --> 00:48:35,160
Speaker 2: and say, do you understand the real harm this site

857
00:48:35,200 --> 00:48:39,600
Speaker 2: is causing? You're a father, And I remember he walked

858
00:48:39,680 --> 00:48:42,200
Speaker 2: right into the hospital. He wouldn't say a word to me,

859
00:48:42,239 --> 00:48:44,960
Speaker 2: and the doors closed, and I remember thinking at that point, well,

860
00:48:44,960 --> 00:48:46,800
Speaker 2: we're gonna have to go talk about this to everyone

861
00:48:46,880 --> 00:48:49,480
Speaker 2: who will listen to us, and that has to change

862
00:48:49,480 --> 00:48:51,520
Speaker 2: because we have to change culture as a part of this.

863
00:48:51,520 --> 00:48:53,040
Speaker 1: This can't be where the story ends.

864
00:48:53,080 --> 00:48:55,799
Speaker 2: He can't just take himself off social media, which he

865
00:48:55,880 --> 00:48:58,839
Speaker 2: did and continue running this site. And so a lot

866
00:48:58,920 --> 00:49:01,520
Speaker 2: of things happened, which I think, think we're really exciting.

867
00:49:01,600 --> 00:49:03,560
Speaker 2: I mean we I think were one of the first

868
00:49:03,719 --> 00:49:07,240
Speaker 2: to show up and threaten his anonymity, so you cannot

869
00:49:07,360 --> 00:49:09,839
Speaker 2: do this type of abuse and remain anonymous.

870
00:49:09,840 --> 00:49:12,160
Speaker 1: So that was friction. That exposure was friction.

871
00:49:12,680 --> 00:49:15,160
Speaker 2: A laws started changing when in the UK they started

872
00:49:15,200 --> 00:49:17,920
Speaker 2: talking about different types of laws that would make it

873
00:49:17,960 --> 00:49:20,319
Speaker 2: harder for folks to access mister deep Fix. They caught

874
00:49:20,320 --> 00:49:23,080
Speaker 2: off access to folks in the UK. That created friction.

875
00:49:23,760 --> 00:49:26,439
Speaker 2: A very valuable server went down for them to take

876
00:49:26,480 --> 00:49:30,680
Speaker 2: it down. At past the CBC published their investigation as well.

877
00:49:31,040 --> 00:49:34,400
Speaker 2: Mister deep Fix finally went offline for good. And I

878
00:49:34,440 --> 00:49:37,200
Speaker 2: think it was such a great lesson in all of

879
00:49:37,239 --> 00:49:41,440
Speaker 2: these players that actually impact how we build out a

880
00:49:41,480 --> 00:49:44,840
Speaker 2: better future. And the headline is friction. It just created

881
00:49:44,920 --> 00:49:48,200
Speaker 2: more and more friction. Now what's happened to him, I

882
00:49:48,239 --> 00:49:51,120
Speaker 2: think is actually is just as important of a headline,

883
00:49:51,120 --> 00:49:53,919
Speaker 2: which is nothing really. I mean, he lost his job

884
00:49:54,440 --> 00:49:58,800
Speaker 2: at the pharmacy. But we've talked to many folks law enforcement.

885
00:49:58,840 --> 00:50:02,040
Speaker 2: I actually spoke to a couple different folks in Canada.

886
00:50:02,040 --> 00:50:05,280
Speaker 2: From the law enforcement perspective, there's no open investigations into

887
00:50:05,600 --> 00:50:08,839
Speaker 2: David Doe. That's his name, and there's thoughts that maybe

888
00:50:08,880 --> 00:50:11,520
Speaker 2: he's left the country. But that's it and I think

889
00:50:11,560 --> 00:50:14,839
Speaker 2: that in and of itself tells its own story absolutely.

890
00:50:14,960 --> 00:50:17,840
Speaker 3: So the last question dve Forne obviously isn't going away.

891
00:50:18,000 --> 00:50:21,200
Speaker 3: So what's the best possible outcome each of you, what

892
00:50:21,280 --> 00:50:23,799
Speaker 3: is your most optimistic case for how it pans out

893
00:50:23,800 --> 00:50:26,480
Speaker 3: in the next five years and how we get there. Carrie,

894
00:50:26,600 --> 00:50:29,960
Speaker 3: you go first, and then VS and Laurie you get

895
00:50:29,960 --> 00:50:31,040
Speaker 3: the last word. Carrie.

896
00:50:31,160 --> 00:50:33,560
Speaker 5: Okay, Well it should probably come as no shock too

897
00:50:33,719 --> 00:50:36,920
Speaker 5: that my best case outcome is that the companies that

898
00:50:37,480 --> 00:50:42,319
Speaker 5: unleashed these harms and profited significantly from them should have

899
00:50:42,400 --> 00:50:46,600
Speaker 5: to pay and there should be disgorgement and compensation for

900
00:50:46,680 --> 00:50:50,279
Speaker 5: all the victims, and it should be financially penalizing for

901
00:50:50,360 --> 00:50:51,120
Speaker 5: these companies.

902
00:50:51,760 --> 00:50:54,520
Speaker 3: So that's would be your best And is that a

903
00:50:54,760 --> 00:50:57,600
Speaker 3: possible outcome? Do you feel like that's a possible outcome.

904
00:50:58,760 --> 00:51:01,919
Speaker 5: I mean, we're on that road with litigation. We'll see

905
00:51:01,920 --> 00:51:04,920
Speaker 5: how it works. I also, you know, I do condemn

906
00:51:05,520 --> 00:51:08,759
Speaker 5: the platforms like the App Store and Google Play who

907
00:51:09,360 --> 00:51:15,080
Speaker 5: for many many years were offering and selling newdifying apps

908
00:51:15,440 --> 00:51:18,719
Speaker 5: and they just kind of were ostriches with their head

909
00:51:18,719 --> 00:51:21,399
Speaker 5: in the sand, pretending like they didn't think these things

910
00:51:21,400 --> 00:51:25,239
Speaker 5: were harmful. I'm really sick and tired of platforms not

911
00:51:25,400 --> 00:51:29,040
Speaker 5: doing anything until something becomes actually criminal. You know, they

912
00:51:29,080 --> 00:51:33,280
Speaker 5: deny that it's harmful until there's an actual criminal law VS.

913
00:51:34,160 --> 00:51:36,920
Speaker 4: You know, that's what I would like in place, some

914
00:51:37,120 --> 00:51:40,880
Speaker 4: kind of criminal law that says, well, look, here are

915
00:51:40,920 --> 00:51:45,480
Speaker 4: the kinds of behaviors, and even intended behaviors are intended

916
00:51:45,520 --> 00:51:49,040
Speaker 4: outcomes that should be criminalized. Here are the penalties that

917
00:51:49,080 --> 00:51:52,520
Speaker 4: you will face if you engage in these acts. And

918
00:51:52,960 --> 00:51:56,000
Speaker 4: this needs to be something that's at least in the

919
00:51:56,120 --> 00:52:00,200
Speaker 4: US nationwide. But more importantly, many of these actors are

920
00:52:00,239 --> 00:52:03,239
Speaker 4: going to be overseas. And so what we need is

921
00:52:03,719 --> 00:52:07,239
Speaker 4: not just US law, but law in various countries and

922
00:52:07,320 --> 00:52:12,640
Speaker 4: some broad international agreements, perhaps enforced by an organization like Interpol,

923
00:52:13,040 --> 00:52:17,160
Speaker 4: which polices things like child grafficking already, they should be

924
00:52:17,200 --> 00:52:19,920
Speaker 4: authorized to do and investigate things like this.

925
00:52:20,480 --> 00:52:23,160
Speaker 3: That's a very good point. That's exactly where it goes

926
00:52:23,200 --> 00:52:25,920
Speaker 3: as broad is what happens is immediately it goes abroad.

927
00:52:26,239 --> 00:52:29,239
Speaker 2: Laurie, last word, Look, I agree with what everyone here

928
00:52:29,280 --> 00:52:31,200
Speaker 2: has said, So I will end with saying, you know,

929
00:52:31,280 --> 00:52:33,560
Speaker 2: my gripe with Silicon Valley, and I have a lot

930
00:52:33,600 --> 00:52:37,120
Speaker 2: of gripes with Silicon Valley, is they're sitting there talking

931
00:52:37,280 --> 00:52:41,279
Speaker 2: about AGI and what happens when AI becomes smarter than us,

932
00:52:41,680 --> 00:52:44,520
Speaker 2: and it doesn't matter, It really doesn't matter, because here's

933
00:52:44,520 --> 00:52:45,680
Speaker 2: what's happening.

934
00:52:45,280 --> 00:52:46,040
Speaker 1: Here on Earth.

935
00:52:46,480 --> 00:52:49,640
Speaker 2: Over here in reality, we have children who are ending

936
00:52:49,680 --> 00:52:52,680
Speaker 2: their lives because of deep fake sex stortion, and people

937
00:52:52,719 --> 00:52:56,160
Speaker 2: are sleeping on it. My hope is that I don't

938
00:52:56,160 --> 00:52:58,799
Speaker 2: have to shout into the ether for years and say

939
00:52:58,800 --> 00:53:00,600
Speaker 2: this is real harm, this is real home. That we

940
00:53:00,640 --> 00:53:03,799
Speaker 2: can get better education, that we can change culture to

941
00:53:04,040 --> 00:53:08,799
Speaker 2: understand that when new technology comes quickly without the correct guardrails,

942
00:53:09,040 --> 00:53:12,680
Speaker 2: the folks who are impacted oftentimes are the ones that

943
00:53:12,760 --> 00:53:15,360
Speaker 2: don't have the voice, and we can make some changes

944
00:53:15,400 --> 00:53:18,000
Speaker 2: from there. No, it doesn't take, to Carrie's point, Oftentimes

945
00:53:18,040 --> 00:53:20,759
Speaker 2: it takes criminal law if I'm putting sitting in my seat.

946
00:53:20,840 --> 00:53:24,279
Speaker 2: Oftentimes it takes public pressure and investigations. I'd love to

947
00:53:24,320 --> 00:53:26,520
Speaker 2: get to a world where Silicon Valley maybe has some

948
00:53:26,600 --> 00:53:29,800
Speaker 2: different folks in it. And this isn't the last thought

949
00:53:29,920 --> 00:53:32,959
Speaker 2: as people are having dinner and talking about AGI after

950
00:53:33,000 --> 00:53:34,160
Speaker 2: taking a cold plunge.

951
00:53:34,239 --> 00:53:34,640
Speaker 1: That's it.

952
00:53:34,920 --> 00:53:36,520
Speaker 3: That's a very good point. But you're going to wait

953
00:53:36,560 --> 00:53:38,800
Speaker 3: a long time for that. Just thank you, But I

954
00:53:38,880 --> 00:53:45,200
Speaker 3: got manifested, go ahead, all right? Sure? Why not? Why not? Anyway?

955
00:53:45,239 --> 00:53:47,120
Speaker 3: I really appreciate it, all three of you. It's a

956
00:53:47,160 --> 00:53:49,080
Speaker 3: really important talk, and we're going to keep coming back

957
00:53:49,080 --> 00:53:51,480
Speaker 3: to it again and again and hopefully see something happen

958
00:53:51,719 --> 00:53:55,359
Speaker 3: by raising awareness of it. For sure. Thank you so much, all.

959
00:53:55,280 --> 00:53:57,960
Speaker 4: Three of you, Thank you, Ker pleasure to be here.

960
00:53:57,840 --> 00:54:01,600
Speaker 3: Thank you so much. Thanks. We've reached out to XAI

961
00:54:01,640 --> 00:54:06,439
Speaker 3: and x for comment. They did not respond. Today's show

962
00:54:06,480 --> 00:54:10,800
Speaker 3: was produced by Michelle Alloy, Katherine Millsop, Madeline LaPlante, Doobie,

963
00:54:10,920 --> 00:54:15,160
Speaker 3: and Kaylin Lynch. Special thanks to Lissa Soap, Anika Robbins

964
00:54:15,239 --> 00:54:18,920
Speaker 3: and Julia Sharp Levine. Our engineers are Fernando Aruda and

965
00:54:19,000 --> 00:54:22,760
Speaker 3: Rick Kwan, and our theme music is by Trackademics. Nishat

966
00:54:22,840 --> 00:54:26,880
Speaker 3: Kerwa is Vox Media's executive producer of podcasts. Go Wherever

967
00:54:26,960 --> 00:54:29,400
Speaker 3: you listen to podcasts, search for On with Kara Swisher

968
00:54:29,400 --> 00:54:31,919
Speaker 3: and hit follow. Thanks for listening to On with Kara

969
00:54:32,000 --> 00:54:34,800
Speaker 3: Swisher from Podium Media, New York Magazine, the vox Media

970
00:54:34,880 --> 00:54:38,279
Speaker 3: Podcast Network and US. We'll be back on Monday with

971
00:54:38,440 --> 00:54:38,680
Speaker 3: more