1
00:00:04,160 --> 00:00:05,280
Speaker 1: There Are No Girls on the Internet.

2
00:00:05,320 --> 00:00:12,680
Speaker 2: As a production of iHeartRadio and Unbossed Creative, I'm Bridge

3
00:00:12,760 --> 00:00:16,240
Speaker 2: Tad and this is there are No Girls on the Internet.

4
00:00:17,800 --> 00:00:19,639
Speaker 2: So I wanted to do a little bit of a

5
00:00:19,680 --> 00:00:22,279
Speaker 2: deeper dive update on a story that we talked about

6
00:00:22,360 --> 00:00:26,640
Speaker 2: in last week's news roundup those awful AI faked images

7
00:00:26,680 --> 00:00:29,680
Speaker 2: of Taylor Swift that were flooding social media last week,

8
00:00:29,920 --> 00:00:32,479
Speaker 2: and just generally get into some of the discourse happening

9
00:00:32,520 --> 00:00:35,400
Speaker 2: around Taylor Swift online because I think it's really telling.

10
00:00:35,800 --> 00:00:37,680
Speaker 2: So we've talked about it on the show before. But

11
00:00:38,080 --> 00:00:43,120
Speaker 2: deep fakes are non consensual AI generated sexually explicit images,

12
00:00:43,400 --> 00:00:46,960
Speaker 2: and they are a growing problem. According to research from

13
00:00:46,960 --> 00:00:51,000
Speaker 2: independent analyst Genevieve O, the number of new pornographic deep

14
00:00:51,040 --> 00:00:55,120
Speaker 2: fake videos has already surged more than ninefold since twenty twenty.

15
00:00:55,440 --> 00:00:57,960
Speaker 2: At the end of last year, the top ten sites

16
00:00:58,000 --> 00:01:01,360
Speaker 2: offering this AI generated deep fake cut hosted more than

17
00:01:01,400 --> 00:01:05,040
Speaker 2: one hundred and fourteen thousand videos, and Tara Swift has

18
00:01:05,120 --> 00:01:08,080
Speaker 2: always been a common target. In fact, on one of

19
00:01:08,120 --> 00:01:11,520
Speaker 2: these AI platforms, she was the first celebrity targeted for

20
00:01:11,640 --> 00:01:15,400
Speaker 2: deep faked AI images. And the reality is this is

21
00:01:15,440 --> 00:01:18,600
Speaker 2: a gender justice issue. It's an issue that impacts women.

22
00:01:18,959 --> 00:01:21,760
Speaker 2: According to a study from deep Trace Labs, ninety six

23
00:01:21,800 --> 00:01:26,120
Speaker 2: percent of deep fakes are non consensual sexual depictions of women.

24
00:01:26,800 --> 00:01:29,720
Speaker 2: So we're calling these images deep fakes, which is not wrong.

25
00:01:29,959 --> 00:01:32,920
Speaker 2: But one technical thing that is important to understand about

26
00:01:32,959 --> 00:01:35,160
Speaker 2: them is that we used to use the term deep

27
00:01:35,200 --> 00:01:39,360
Speaker 2: fakes to describe content where someone's head or was put

28
00:01:39,400 --> 00:01:42,480
Speaker 2: on another person's body to give the impression that that

29
00:01:42,680 --> 00:01:46,360
Speaker 2: person is nude or doing something sexual explicit. These terror

30
00:01:46,400 --> 00:01:49,120
Speaker 2: Swift deep fakes are different in that they were created

31
00:01:49,320 --> 00:01:53,440
Speaker 2: out of whole cloth using commercial AI tools. The images

32
00:01:53,520 --> 00:01:57,280
Speaker 2: originated on the alternative messaging platform Telegram, then made their

33
00:01:57,280 --> 00:02:00,800
Speaker 2: way to fourtune. This Telegram group was actually kind of

34
00:02:00,840 --> 00:02:04,520
Speaker 2: a marketplace for celebrity deep fakes, where users were requesting

35
00:02:04,560 --> 00:02:09,200
Speaker 2: and trading deep faked images of specific celebrities since last year. Now,

36
00:02:09,400 --> 00:02:13,760
Speaker 2: most commercial AI image generators do have some safeguards in

37
00:02:13,800 --> 00:02:16,160
Speaker 2: place that are meant to prevent abuse. So if you

38
00:02:16,240 --> 00:02:19,799
Speaker 2: ask AI to generate an image that depicts a sexualized

39
00:02:19,800 --> 00:02:22,200
Speaker 2: image of a child, for instance, it is not supposed

40
00:02:22,200 --> 00:02:25,280
Speaker 2: to do that, but people are able to circumvent the

41
00:02:25,320 --> 00:02:28,560
Speaker 2: moderation to generate non consensual deep fake images like the

42
00:02:28,600 --> 00:02:31,800
Speaker 2: ones we saw Taylor Swift pretty easily. I have to

43
00:02:31,800 --> 00:02:34,240
Speaker 2: give a major shout out to four h four media

44
00:02:34,360 --> 00:02:37,200
Speaker 2: right here because they did a great job of digging

45
00:02:37,520 --> 00:02:40,400
Speaker 2: and found that those images of Taylor Swift were likely

46
00:02:40,440 --> 00:02:45,000
Speaker 2: made on Microsoft's AI creation tool called Designer, So Designer

47
00:02:45,200 --> 00:02:48,359
Speaker 2: does have some safeguards in place to prevent the platform

48
00:02:48,400 --> 00:02:51,680
Speaker 2: from being used to create this kind of content of celebrities. However,

49
00:02:52,200 --> 00:02:54,360
Speaker 2: four for media ran a couple of tests and found

50
00:02:54,360 --> 00:02:56,600
Speaker 2: that Designer will not generate an image if you just

51
00:02:56,639 --> 00:03:00,480
Speaker 2: put Jennifer Aniston, but they were able to generate suggestive

52
00:03:00,520 --> 00:03:03,919
Speaker 2: images of Jennifer Aniston if you use the phrase Jennifer

53
00:03:04,040 --> 00:03:06,959
Speaker 2: actor Aniston. So just put actor in between Jennifer and

54
00:03:07,000 --> 00:03:11,000
Speaker 2: Aniston and boom, non consensual deep fakes. So, prior to

55
00:03:11,040 --> 00:03:14,000
Speaker 2: these deep fake images of Taylor Swift going viral on Twitter,

56
00:03:14,280 --> 00:03:17,680
Speaker 2: a user in that telegram group recommended that members who

57
00:03:17,680 --> 00:03:20,480
Speaker 2: were trying to make sexual explicit images of Taylor Swift

58
00:03:21,040 --> 00:03:24,520
Speaker 2: use the phrase Taylor singer Swift instead of just Taylor

59
00:03:24,560 --> 00:03:28,320
Speaker 2: Swift to generate those images. Basically, anyone can get around

60
00:03:28,360 --> 00:03:32,120
Speaker 2: these guardrails pretty easily. They're full of loopholes, and to

61
00:03:32,120 --> 00:03:34,800
Speaker 2: get deep fakes of celebrities engaging in what looks like

62
00:03:34,920 --> 00:03:38,240
Speaker 2: sex acts, all people needed to do was describe activities

63
00:03:38,240 --> 00:03:41,240
Speaker 2: that kind of look like sex ax without using sexual

64
00:03:41,320 --> 00:03:45,520
Speaker 2: explicit terms. Now, none of this is terribly surprising. It's

65
00:03:45,520 --> 00:03:47,560
Speaker 2: the kind of thing that people have been warning is

66
00:03:47,600 --> 00:03:50,480
Speaker 2: really easy to do with AI for a long time.

67
00:03:50,840 --> 00:03:54,000
Speaker 2: I cannot stress how much for for Media stayed on

68
00:03:54,120 --> 00:03:56,200
Speaker 2: Microsoft's neck on this story.

69
00:03:56,600 --> 00:03:57,000
Speaker 1: At first.

70
00:03:57,080 --> 00:03:59,560
Speaker 2: Microsoft was kind of trying to weasel out of this

71
00:03:59,600 --> 00:04:02,520
Speaker 2: and deny that their platform was being used to create

72
00:04:02,520 --> 00:04:05,240
Speaker 2: these images and that their platform is where these tailor

73
00:04:05,320 --> 00:04:09,000
Speaker 2: Swift images originated. They initially said we are investigating these

74
00:04:09,000 --> 00:04:12,080
Speaker 2: reports and are taking appropriate action to address them. Our

75
00:04:12,120 --> 00:04:14,320
Speaker 2: code of conduct prohibits the use of our tools for

76
00:04:14,360 --> 00:04:17,320
Speaker 2: the creation of adult or non consensual intimate content, and

77
00:04:17,400 --> 00:04:20,040
Speaker 2: any repeated attempts to produce content that goes against our

78
00:04:20,080 --> 00:04:22,880
Speaker 2: policies may result in a loss of access to the surface.

79
00:04:23,240 --> 00:04:25,760
Speaker 2: We have large teams working on the development of guardrails

80
00:04:25,760 --> 00:04:28,800
Speaker 2: and other safety systems in line with our responsible AI principles,

81
00:04:28,880 --> 00:04:32,920
Speaker 2: including content filtering, operational monitoring, and abuse detection to mitigate

82
00:04:33,000 --> 00:04:36,440
Speaker 2: misuse of systems and help create a safer environment for users.

83
00:04:36,680 --> 00:04:39,680
Speaker 2: So after that, Microsoft said they did an investigation, but

84
00:04:39,720 --> 00:04:42,360
Speaker 2: they were unable to confirm the images were made using

85
00:04:42,400 --> 00:04:47,880
Speaker 2: their tool. But after Form fours reporting, Microsoft did close

86
00:04:47,920 --> 00:04:49,839
Speaker 2: loopholes that would have allowed for these images to be

87
00:04:49,880 --> 00:04:52,440
Speaker 2: created for and four Media was able to verify that

88
00:04:52,480 --> 00:04:55,719
Speaker 2: those loopholes did work to create these kinds of images

89
00:04:55,760 --> 00:04:59,360
Speaker 2: on Thursday before Microsoft made its changes, and that those

90
00:04:59,360 --> 00:05:02,560
Speaker 2: loopholes stopped working after four or four Media published an

91
00:05:02,640 --> 00:05:05,080
Speaker 2: article on it. So it just really does seem like

92
00:05:05,240 --> 00:05:08,440
Speaker 2: Microsoft was the culprit, and it's a really important reminder

93
00:05:08,520 --> 00:05:11,600
Speaker 2: of the importance of good journalism to keep people accountable

94
00:05:11,640 --> 00:05:13,520
Speaker 2: for the harms that their technology causes.

95
00:05:13,400 --> 00:05:15,599
Speaker 1: Because it would have been really easy.

96
00:05:15,320 --> 00:05:18,360
Speaker 2: For Microsoft to just keep denying that they were responsible

97
00:05:18,360 --> 00:05:21,520
Speaker 2: for these images and just quietly close that loophole that

98
00:05:21,640 --> 00:05:24,480
Speaker 2: allowed for images like the ones of Taylor Swift to

99
00:05:24,480 --> 00:05:27,560
Speaker 2: be made on their platform. After this change, users on

100
00:05:27,600 --> 00:05:30,720
Speaker 2: Telegram were complaining that it seemed like the loophole had

101
00:05:30,760 --> 00:05:33,120
Speaker 2: been fixed that allowed for them to make these celebrity

102
00:05:33,160 --> 00:05:37,000
Speaker 2: deep thinks. But unfortunately that Telegram group where people were

103
00:05:37,120 --> 00:05:41,120
Speaker 2: requesting and trading non consensual deep bake images of celebrities

104
00:05:41,400 --> 00:05:45,200
Speaker 2: is still active because Telegram does not really remove people

105
00:05:45,240 --> 00:05:47,159
Speaker 2: who use their platform to do this kind of thing,

106
00:05:47,440 --> 00:05:51,000
Speaker 2: and people on that platform are already talking about what

107
00:05:51,040 --> 00:05:54,760
Speaker 2: AI tools they're going to use to continue creating non

108
00:05:54,760 --> 00:05:58,120
Speaker 2: consensual deep stak porn of celebrities in the future since

109
00:05:58,200 --> 00:06:01,280
Speaker 2: Microsoft closed the loophole that allowed them to use their platform,

110
00:06:01,800 --> 00:06:03,960
Speaker 2: so they're just going to keep doing this and it

111
00:06:04,040 --> 00:06:07,000
Speaker 2: is shockingly easy for them to keep doing so. These

112
00:06:07,040 --> 00:06:09,599
Speaker 2: tailor Swift images had been on Telegram and four Chan

113
00:06:09,720 --> 00:06:12,960
Speaker 2: for a while, but one specific user who subscribes to

114
00:06:12,960 --> 00:06:15,880
Speaker 2: Twitter blue so has a verified check mark and likely

115
00:06:15,920 --> 00:06:19,560
Speaker 2: has their content algorithmically boosted on Twitter, brought them over

116
00:06:19,600 --> 00:06:23,360
Speaker 2: to Twitter. It took Twitter seventeen hours to suspend that

117
00:06:23,480 --> 00:06:28,320
Speaker 2: user and delete that user's posts with the tailor Swift images. However,

118
00:06:28,720 --> 00:06:31,960
Speaker 2: because of the way the internet works, those images were

119
00:06:32,040 --> 00:06:35,839
Speaker 2: still all over Twitter. Like, just suspending that one user

120
00:06:36,040 --> 00:06:37,680
Speaker 2: does not mean all those images are going to go

121
00:06:37,720 --> 00:06:40,480
Speaker 2: away to So the way that Twitter handled this whole

122
00:06:40,520 --> 00:06:44,400
Speaker 2: thing I think is just abysmal and also really telling. First,

123
00:06:44,480 --> 00:06:47,360
Speaker 2: I guess because those images were really all over Twitter,

124
00:06:47,720 --> 00:06:50,520
Speaker 2: they just tried to fix it by blocking people from

125
00:06:50,560 --> 00:06:54,560
Speaker 2: being able to search Taylor Swift temporarily. Joe Benerach, head

126
00:06:54,560 --> 00:06:57,520
Speaker 2: of Business Operations at Twitter, said, this is a temporary

127
00:06:57,560 --> 00:06:59,640
Speaker 2: action and done with an abundance of caution, as we

128
00:06:59,680 --> 00:07:03,200
Speaker 2: prayor or tie safety on this issue. But because it's Twitter,

129
00:07:03,680 --> 00:07:07,680
Speaker 2: this was actually super easily circumvented. Attempts to search for

130
00:07:07,720 --> 00:07:11,280
Speaker 2: Taylor Swift's name without quotation marks on Twitter on Monday

131
00:07:11,480 --> 00:07:14,280
Speaker 2: resulted in an error message prompting for users to reach

132
00:07:14,320 --> 00:07:16,800
Speaker 2: by their search, which added, don't fret, it's not your thought.

133
00:07:17,080 --> 00:07:17,960
Speaker 1: However, to get.

134
00:07:17,800 --> 00:07:19,840
Speaker 2: Around this, all you would have to do is put

135
00:07:19,920 --> 00:07:22,440
Speaker 2: quotation marks around her name and then you could search

136
00:07:22,440 --> 00:07:25,480
Speaker 2: her name like normal. So, even though this was not

137
00:07:25,720 --> 00:07:28,360
Speaker 2: a very effective way to deal with the issue, I

138
00:07:28,360 --> 00:07:30,920
Speaker 2: would still be taking issue with it had they been

139
00:07:30,960 --> 00:07:33,800
Speaker 2: able to successfully block Taylor Swift from search on Twitter,

140
00:07:34,160 --> 00:07:37,400
Speaker 2: because why should the target of this kind of sexual

141
00:07:37,480 --> 00:07:41,520
Speaker 2: exploitation have to have their name blocked on social media search? Like,

142
00:07:41,600 --> 00:07:43,960
Speaker 2: what if this happened to somebody who is not famous?

143
00:07:44,360 --> 00:07:47,640
Speaker 2: You know, I'm thinking of content creators and journalists, people

144
00:07:47,720 --> 00:07:49,560
Speaker 2: who need to be able to show up on those

145
00:07:49,600 --> 00:07:52,960
Speaker 2: platforms for their careers. It's basically like saying that when

146
00:07:52,960 --> 00:07:55,400
Speaker 2: you are targeted for this kind of thing, the only

147
00:07:55,440 --> 00:07:57,360
Speaker 2: way to keep you safe in these online.

148
00:07:57,040 --> 00:07:59,720
Speaker 1: Spaces is to obscure your presence.

149
00:07:59,280 --> 00:08:02,720
Speaker 2: There, which I take issue with because participation in online

150
00:08:02,720 --> 00:08:06,040
Speaker 2: platforms is increasingly linked to help people just show up

151
00:08:06,040 --> 00:08:08,800
Speaker 2: in twenty twenty four socially, politically.

152
00:08:08,280 --> 00:08:08,920
Speaker 1: And civically.

153
00:08:09,120 --> 00:08:11,880
Speaker 2: So if the response to this is to make targeted people,

154
00:08:12,120 --> 00:08:15,080
Speaker 2: most of whom are women, show up less in those spaces,

155
00:08:15,440 --> 00:08:18,640
Speaker 2: it means that they will be even more marginalized online,

156
00:08:18,720 --> 00:08:22,080
Speaker 2: a space where women are already marginalized. It just feels

157
00:08:22,080 --> 00:08:25,280
Speaker 2: like the fix is to further obscure and silence women's

158
00:08:25,320 --> 00:08:28,640
Speaker 2: voices in a space where we are already obscured and silenced.

159
00:08:28,640 --> 00:08:31,000
Speaker 2: And I don't think it's equitable, and I don't think

160
00:08:31,040 --> 00:08:32,920
Speaker 2: that part of the solution of how we deal with

161
00:08:32,960 --> 00:08:36,000
Speaker 2: this is to make us even more invisible on online spaces.

162
00:08:39,559 --> 00:08:52,240
Speaker 1: Let's take a quick break at our back.

163
00:08:54,520 --> 00:08:56,920
Speaker 2: Elon Musk said that Twitter was going to hire one

164
00:08:57,000 --> 00:09:00,480
Speaker 2: hundred full time staffers to do content moderation to combat

165
00:09:00,520 --> 00:09:03,320
Speaker 2: sexually abusive material on Twitter. He said that he would

166
00:09:03,360 --> 00:09:06,280
Speaker 2: be building a trust and safety center in Austin, Texas

167
00:09:06,600 --> 00:09:09,280
Speaker 2: and higher in house agents who will be tasked with

168
00:09:09,360 --> 00:09:13,320
Speaker 2: enforcing the sites content and safety rules. So I don't

169
00:09:13,320 --> 00:09:15,240
Speaker 2: even have time to get into how annoying this is

170
00:09:15,240 --> 00:09:19,240
Speaker 2: to me because basically, the very same teams that Elon

171
00:09:19,400 --> 00:09:21,760
Speaker 2: fired right after he took over at Twitter in twenty

172
00:09:21,800 --> 00:09:25,079
Speaker 2: twenty two, He's being like, oh, whoops, actually we needed

173
00:09:25,080 --> 00:09:25,520
Speaker 2: those teams.

174
00:09:25,600 --> 00:09:27,240
Speaker 1: Those teams actually did something important.

175
00:09:27,600 --> 00:09:29,720
Speaker 2: So Twitter released this whole thing this week about how

176
00:09:29,760 --> 00:09:33,880
Speaker 2: they were taking really seriously sexually exploitative content, particularly content

177
00:09:33,920 --> 00:09:36,920
Speaker 2: that involved children. And you might remember that, for whatever reason,

178
00:09:36,920 --> 00:09:40,360
Speaker 2: a while back, Elon Musk made combating child sexual abuse

179
00:09:40,400 --> 00:09:42,640
Speaker 2: material like his thing unless he started talking about it

180
00:09:42,640 --> 00:09:46,080
Speaker 2: and grandstanding about how much work he did to fight it. However,

181
00:09:46,559 --> 00:09:49,280
Speaker 2: Wired reported that back in twenty twenty two, when he

182
00:09:49,320 --> 00:09:51,960
Speaker 2: took over at Twitter, even though he said that removing

183
00:09:52,080 --> 00:09:55,920
Speaker 2: child exploitation was his quote number one priority, after all

184
00:09:55,960 --> 00:10:00,640
Speaker 2: the firings and resignations, just one sole staff member remained

185
00:10:00,640 --> 00:10:03,760
Speaker 2: on a key team dedicated to removing child sexual abuse

186
00:10:03,800 --> 00:10:08,040
Speaker 2: content from Twitter. And remember, he also reinstated the account

187
00:10:08,080 --> 00:10:12,080
Speaker 2: of a right wing provocateur who posted child exploitation material

188
00:10:12,160 --> 00:10:15,400
Speaker 2: on Twitter, and they explained that decision to reinstate his

189
00:10:15,400 --> 00:10:19,040
Speaker 2: account by saying, well, if abusive sexual child content is

190
00:10:19,080 --> 00:10:23,120
Speaker 2: posted on Twitter to express outrage, then it's okay. Basically,

191
00:10:23,120 --> 00:10:25,280
Speaker 2: what I'm saying is that what Elon Musk says he

192
00:10:25,320 --> 00:10:28,200
Speaker 2: cares about doing and what he actually does in practice are.

193
00:10:28,080 --> 00:10:29,200
Speaker 1: Two different things.

194
00:10:29,679 --> 00:10:32,240
Speaker 2: You know, it was his choice to gut the trust

195
00:10:32,240 --> 00:10:34,600
Speaker 2: and safety team and content moderation teams when he took

196
00:10:34,600 --> 00:10:38,400
Speaker 2: over at Twitter. So now, in classic Elon form, he's

197
00:10:38,480 --> 00:10:41,600
Speaker 2: backtracking on that great, big mistake. This is actually a

198
00:10:41,600 --> 00:10:44,240
Speaker 2: pattern for Elon Musk. It's like he comes in, gets

199
00:10:44,320 --> 00:10:46,760
Speaker 2: rid of everything that people know that you need to

200
00:10:46,800 --> 00:10:50,040
Speaker 2: make a platform work, realizes, oh crap, turns out there

201
00:10:50,080 --> 00:10:52,160
Speaker 2: was a reason for having that and we actually need it,

202
00:10:52,200 --> 00:10:54,880
Speaker 2: reinstates it, and then I guess just like stands back

203
00:10:54,920 --> 00:10:57,640
Speaker 2: and calls that innovation. When he first took over at Twitter,

204
00:10:57,720 --> 00:11:00,760
Speaker 2: he immediately let people who had been previously spended for

205
00:11:01,040 --> 00:11:04,920
Speaker 2: abusive content or hate speech back on the platform, people

206
00:11:05,080 --> 00:11:08,679
Speaker 2: like Kanye West, who Elon Musk then had to resuspend

207
00:11:09,000 --> 00:11:11,559
Speaker 2: after Kanye West tweeted out of swastika.

208
00:11:11,720 --> 00:11:13,600
Speaker 1: So this is just what Elon Musk does.

209
00:11:14,200 --> 00:11:18,080
Speaker 2: I just can't believe he would announce creating content moderation

210
00:11:18,240 --> 00:11:21,520
Speaker 2: teams and trust in safety teams, like it's some new

211
00:11:21,800 --> 00:11:24,360
Speaker 2: innovation that he has come up with as opposed to

212
00:11:24,440 --> 00:11:27,080
Speaker 2: something that was already in place at Twitter that he

213
00:11:27,280 --> 00:11:30,079
Speaker 2: got rid of and now is reinstating. And it's funny

214
00:11:30,080 --> 00:11:32,280
Speaker 2: because I was talking to producer Mike about this, and

215
00:11:32,320 --> 00:11:34,400
Speaker 2: he made a very good point, which is that all

216
00:11:34,440 --> 00:11:37,240
Speaker 2: of this assumes that Elon will actually follow his word

217
00:11:37,520 --> 00:11:39,320
Speaker 2: and do the thing that he says he's going to do.

218
00:11:39,400 --> 00:11:43,040
Speaker 2: So will he actually create this content moderation team, will

219
00:11:43,080 --> 00:11:46,720
Speaker 2: he actually create a trust and safety team? Who knows?

220
00:11:46,760 --> 00:11:50,080
Speaker 2: Because Elon has such a poor track record of actually

221
00:11:50,280 --> 00:11:52,240
Speaker 2: doing the things he promises to do.

222
00:11:52,760 --> 00:11:52,880
Speaker 1: So.

223
00:11:53,120 --> 00:11:56,280
Speaker 2: There's been this huge reaction about these Taylor Swift deep fakes,

224
00:11:56,280 --> 00:11:58,480
Speaker 2: and rightly so, but they are not the first time

225
00:11:58,520 --> 00:12:00,880
Speaker 2: this has happened. We've covered how this kind of thing

226
00:12:00,920 --> 00:12:03,280
Speaker 2: is happening at schools across the country in places like

227
00:12:03,320 --> 00:12:05,800
Speaker 2: New Jersey and Colorado, and how they've been dealing with

228
00:12:05,880 --> 00:12:10,080
Speaker 2: groups of male students essentially creating ai deep fake extortion

229
00:12:10,240 --> 00:12:13,439
Speaker 2: rings of the girls in their school, where the boys

230
00:12:13,480 --> 00:12:16,000
Speaker 2: in the school are invited to pay to see deep

231
00:12:16,040 --> 00:12:19,160
Speaker 2: fakes nude images of their classmates, and the girls are

232
00:12:19,240 --> 00:12:23,000
Speaker 2: basically forced into being extorted to pay money to avoid

233
00:12:23,040 --> 00:12:26,840
Speaker 2: their images being circulated. In New Jersey, high school student

234
00:12:26,960 --> 00:12:30,880
Speaker 2: Francesca Moni, a teenage victim of non consensual sexually explicit

235
00:12:30,960 --> 00:12:34,560
Speaker 2: deep fakes, joined Representative Joe Morel to tell her story

236
00:12:34,559 --> 00:12:36,680
Speaker 2: and advocate for legislation that would make this kind of

237
00:12:36,679 --> 00:12:40,680
Speaker 2: thing a crime. And teen marvelstar Social Gomez was targeted

238
00:12:40,720 --> 00:12:44,040
Speaker 2: with non consensual deep fakes earlier this year, and she

239
00:12:44,200 --> 00:12:48,520
Speaker 2: is only seventeen, a minor. It's really heartbreaking because she

240
00:12:48,679 --> 00:12:51,440
Speaker 2: says that Twitter pretty much did nothing to take these

241
00:12:51,480 --> 00:12:55,200
Speaker 2: images down, they did not respond to requests for comment

242
00:12:55,240 --> 00:12:57,080
Speaker 2: about it, and that she said in a podcast that

243
00:12:57,120 --> 00:13:00,600
Speaker 2: she basically had to make personal peace with the fact

244
00:13:00,640 --> 00:13:02,359
Speaker 2: that these deep faked images.

245
00:13:01,960 --> 00:13:02,880
Speaker 1: Of her are out there.

246
00:13:03,280 --> 00:13:07,360
Speaker 2: Like, imagine what we are asking young people, children minors

247
00:13:07,600 --> 00:13:11,079
Speaker 2: to make individual peace with because the powers that be

248
00:13:11,440 --> 00:13:14,280
Speaker 2: are not taking it seriously or doing anything to prevent

249
00:13:14,280 --> 00:13:16,559
Speaker 2: this kind of thing from happening or dealing with it

250
00:13:16,600 --> 00:13:17,400
Speaker 2: when it does happen.

251
00:13:17,920 --> 00:13:18,600
Speaker 1: And I do want to.

252
00:13:18,559 --> 00:13:21,080
Speaker 2: Say something else about Social Gomes. We've talked about this

253
00:13:21,120 --> 00:13:23,319
Speaker 2: a bit on the show before. But there does seem

254
00:13:23,360 --> 00:13:25,920
Speaker 2: to be some kind of a teared system of response

255
00:13:26,280 --> 00:13:30,040
Speaker 2: depending on how famous the target is, because certainly there

256
00:13:30,120 --> 00:13:33,200
Speaker 2: was not this kind of outcry for social gomes who

257
00:13:33,240 --> 00:13:36,160
Speaker 2: was targeted for AI deep think images before Taylor Swift.

258
00:13:36,559 --> 00:13:37,439
Speaker 1: We kind of saw.

259
00:13:37,280 --> 00:13:40,960
Speaker 2: Something similar under the earlier iteration of Twitter, where marginalized

260
00:13:40,960 --> 00:13:44,120
Speaker 2: people were speaking up about getting torrents of online harassments

261
00:13:44,160 --> 00:13:47,440
Speaker 2: and abuse, like black women activists were really open about

262
00:13:47,440 --> 00:13:50,520
Speaker 2: what they were experiencing. However, it took Saturday Night Live

263
00:13:50,600 --> 00:13:53,720
Speaker 2: actor Leslie Jones to speak up on Twitter when she

264
00:13:53,960 --> 00:13:57,199
Speaker 2: was facing a torrent of harassment after her Ghostbuster's remake

265
00:13:57,240 --> 00:14:00,480
Speaker 2: came out that Twitter took any meaningful steps to suspend

266
00:14:00,520 --> 00:14:02,920
Speaker 2: the accounts of people who were doing it. Then Twitter

267
00:14:03,000 --> 00:14:07,360
Speaker 2: CEO Jack Dorsey personally intervened to suspend those accounts, and

268
00:14:07,440 --> 00:14:09,960
Speaker 2: like replied to Leslie Jones saying that he was sorry.

269
00:14:09,760 --> 00:14:10,360
Speaker 1: That had happened.

270
00:14:10,600 --> 00:14:13,760
Speaker 2: But our teens who are not famous are already being

271
00:14:13,840 --> 00:14:16,599
Speaker 2: targeted for non consensual deep fake images. You know, you

272
00:14:16,640 --> 00:14:19,680
Speaker 2: shouldn't have to be Taylor Swift level of famous to

273
00:14:19,720 --> 00:14:22,520
Speaker 2: get people with power to take this seriously. When we

274
00:14:22,560 --> 00:14:25,400
Speaker 2: are all at risk now. The Taylor Swift deep fake

275
00:14:25,440 --> 00:14:28,800
Speaker 2: images on Twitter have triggered a response from the Biden administration.

276
00:14:29,160 --> 00:14:31,400
Speaker 2: At a White House press briefing, a reporter asked if

277
00:14:31,400 --> 00:14:35,560
Speaker 2: Biden supported legislation to ban AI generated porn images, and

278
00:14:35,640 --> 00:14:38,720
Speaker 2: White House Press Secretary Kareem Jean Pierre said, we are

279
00:14:38,760 --> 00:14:41,480
Speaker 2: alarmed by the reports of the circulation of false images.

280
00:14:41,640 --> 00:14:43,000
Speaker 2: We are going to do what we can to deal

281
00:14:43,000 --> 00:14:46,040
Speaker 2: with the issue. The Screen Actors Guild, the union that

282
00:14:46,120 --> 00:14:49,680
Speaker 2: represents performers in Hollywood, has spoken up to saying the

283
00:14:49,680 --> 00:14:53,480
Speaker 2: development and dissemination of fake images, especially those of lewd nature,

284
00:14:53,640 --> 00:14:55,920
Speaker 2: without someone's consent, must be made illegal.

285
00:14:56,120 --> 00:14:57,120
Speaker 1: As a society, we.

286
00:14:57,120 --> 00:14:59,360
Speaker 2: Have it in our power to control these technologies, but

287
00:14:59,400 --> 00:15:01,560
Speaker 2: we must act now before it is too late. The

288
00:15:01,560 --> 00:15:04,600
Speaker 2: way performers are depicted in AI without their consent has

289
00:15:04,640 --> 00:15:07,760
Speaker 2: been a major point of contention with actors' unions like SAG.

290
00:15:08,000 --> 00:15:11,560
Speaker 2: The union also supports broader legislation like the No AI

291
00:15:11,720 --> 00:15:14,360
Speaker 2: Frauds Act that would prohibit the use of AI to

292
00:15:14,440 --> 00:15:18,200
Speaker 2: distribute faked replicas of real people without their consent, regardless

293
00:15:18,240 --> 00:15:21,080
Speaker 2: of whether those images are sexually explicit or not. SAG

294
00:15:21,120 --> 00:15:24,160
Speaker 2: went on to say we support Taylor and women everywhere

295
00:15:24,160 --> 00:15:26,000
Speaker 2: who are the victims of this kind of theft of

296
00:15:26,040 --> 00:15:29,000
Speaker 2: their privacy and right to autonomy. So I do have

297
00:15:29,080 --> 00:15:33,120
Speaker 2: to add that there is something weird happening around Taylor

298
00:15:33,160 --> 00:15:35,640
Speaker 2: Swift right now, Like it does not surprise me that

299
00:15:35,680 --> 00:15:39,440
Speaker 2: these deep fakes were first circulated on Telegram and four Chan,

300
00:15:39,760 --> 00:15:44,280
Speaker 2: where far right extremists proliferate. So just yesterday, failed Republican

301
00:15:44,320 --> 00:15:49,560
Speaker 2: presidential candidate Viviak Ramaswami spewed some pretty wacky conspiracy theories

302
00:15:49,680 --> 00:15:53,320
Speaker 2: that Taylor Swift and her boyfriend NFL football player Travis

303
00:15:53,360 --> 00:15:57,280
Speaker 2: Kelchi were being like artificially culturally propped up by the

304
00:15:57,320 --> 00:16:00,920
Speaker 2: media for some kind of a democratic, I don't know,

305
00:16:01,000 --> 00:16:04,560
Speaker 2: psyop campaign. He tweeted, I wonder who's going to win

306
00:16:04,600 --> 00:16:06,880
Speaker 2: the Super Bowl next month, and I wonder if there's

307
00:16:06,920 --> 00:16:10,720
Speaker 2: a major presidential endorsement coming from an artificially culturally propped

308
00:16:10,760 --> 00:16:14,880
Speaker 2: up couple this fall. Taylor Swift's boyfriend plays for the Chiefs,

309
00:16:14,920 --> 00:16:16,920
Speaker 2: who just won the playoff game and are going.

310
00:16:16,800 --> 00:16:19,040
Speaker 1: To the super Bowl. And on Fox News this.

311
00:16:19,040 --> 00:16:23,400
Speaker 2: Week they called her boyfriend mister Pfizer because he promoted

312
00:16:23,440 --> 00:16:27,320
Speaker 2: vaccination to curb COVID. So something I've always found weird

313
00:16:27,320 --> 00:16:30,760
Speaker 2: about this is that Taylor Swift is not even super

314
00:16:31,200 --> 00:16:33,480
Speaker 2: publicly or explicitly that political.

315
00:16:33,800 --> 00:16:35,960
Speaker 1: Like she did endorse Biden in twenty twenty.

316
00:16:36,120 --> 00:16:38,360
Speaker 2: She's done a few political things here or there, but

317
00:16:38,440 --> 00:16:41,440
Speaker 2: she's not some like radical political figure. She didn't even

318
00:16:41,640 --> 00:16:44,400
Speaker 2: refer to herself as a feminist until pretty recently. So, like,

319
00:16:44,520 --> 00:16:47,760
Speaker 2: what gives my theory is that I think that Taylor

320
00:16:47,800 --> 00:16:51,080
Speaker 2: Swift represents the anxiety that the right has around young,

321
00:16:51,280 --> 00:16:52,360
Speaker 2: childless white women.

322
00:16:52,480 --> 00:16:52,880
Speaker 1: Right now.

323
00:16:53,200 --> 00:16:55,200
Speaker 2: We know that white women who are married to white

324
00:16:55,280 --> 00:16:58,920
Speaker 2: men tend to vote Republican demographically speaking, like they went

325
00:16:59,120 --> 00:17:02,000
Speaker 2: really hard for Trump, but that it's not so for

326
00:17:02,080 --> 00:17:04,600
Speaker 2: their unmarried white women counterparts.

327
00:17:04,880 --> 00:17:07,040
Speaker 1: And I think the Republican Party just.

328
00:17:07,000 --> 00:17:09,439
Speaker 2: Does not have a lot to offer these women, and

329
00:17:09,480 --> 00:17:12,000
Speaker 2: they know that, so instead of even trying to offer

330
00:17:12,040 --> 00:17:15,200
Speaker 2: them anything, they just lash out at Taylor Swift, who,

331
00:17:15,240 --> 00:17:18,840
Speaker 2: for whatever reason they have decided is representative of single,

332
00:17:19,000 --> 00:17:22,280
Speaker 2: unmarried white women because she is so visible and so powerful.

333
00:17:22,640 --> 00:17:25,000
Speaker 2: And I think that is what they really hate about

334
00:17:25,000 --> 00:17:27,800
Speaker 2: Taylor Swift. That is why she is a target, because

335
00:17:27,800 --> 00:17:31,040
Speaker 2: they hate her visibility. They hate that the cameras cut

336
00:17:31,080 --> 00:17:33,160
Speaker 2: to her at NFL games when she's in the stands

337
00:17:33,400 --> 00:17:36,280
Speaker 2: as if she controls the cameras. They hate that she

338
00:17:36,359 --> 00:17:39,080
Speaker 2: is not acting ashamed to be young and unmarried and

339
00:17:39,160 --> 00:17:42,359
Speaker 2: having fun and having power. So targeting her with deep

340
00:17:42,400 --> 00:17:45,240
Speaker 2: figs is a way to sexually humiliate and humble her,

341
00:17:45,560 --> 00:17:47,919
Speaker 2: to take away that power, and to remind her that

342
00:17:47,960 --> 00:17:50,080
Speaker 2: at the end of the day, we can still reduce

343
00:17:50,119 --> 00:17:52,919
Speaker 2: you to a sexualized object whenever we want, and we

344
00:17:53,000 --> 00:17:54,920
Speaker 2: have the tech to do it. And I remember when

345
00:17:54,920 --> 00:17:58,640
Speaker 2: the conversation was around celebrity images being hacked or stolen

346
00:17:58,680 --> 00:18:01,240
Speaker 2: and released to the public, people would say like, oh, well,

347
00:18:01,280 --> 00:18:03,160
Speaker 2: just don't take those kind of pictures if you don't

348
00:18:03,160 --> 00:18:05,879
Speaker 2: want other people to see them. But now that anyone

349
00:18:05,960 --> 00:18:09,360
Speaker 2: can be targeted for non consensual deep fake sexualized images,

350
00:18:09,720 --> 00:18:12,720
Speaker 2: what will the guidance be then, Like don't be too visible,

351
00:18:12,880 --> 00:18:15,840
Speaker 2: don't be annoying, don't be a leftist, don't be a woman,

352
00:18:16,000 --> 00:18:18,800
Speaker 2: don't exist. And I think that is really the question

353
00:18:18,880 --> 00:18:20,800
Speaker 2: we need to grapple with. Do we want to live

354
00:18:20,800 --> 00:18:24,159
Speaker 2: in a future where any woman, any girl, any person

355
00:18:24,240 --> 00:18:27,879
Speaker 2: can be reduced to a sexual object via commercially available

356
00:18:27,920 --> 00:18:32,800
Speaker 2: AI platforms that someone like Microsoft is making money off of.

357
00:18:33,440 --> 00:18:36,320
Speaker 2: I know it is grim and on the Patreon a

358
00:18:36,359 --> 00:18:39,399
Speaker 2: listener Sierra actually pointed something out that this is an

359
00:18:39,440 --> 00:18:41,840
Speaker 2: issue that is frustrating because it feels like there's not

360
00:18:41,920 --> 00:18:44,159
Speaker 2: a ton that we can do here in the United States.

361
00:18:44,200 --> 00:18:46,959
Speaker 2: We're sort of waiting and watching as lawmakers direct their

362
00:18:47,000 --> 00:18:49,400
Speaker 2: feet and kind of kick the can and pass the buck,

363
00:18:49,680 --> 00:18:53,040
Speaker 2: but there is still no federal legislation prohibiting someone from

364
00:18:53,040 --> 00:18:55,800
Speaker 2: making deep fake images. I would say that one thing

365
00:18:55,800 --> 00:18:58,000
Speaker 2: that we can all do right now is kind of cultural.

366
00:18:58,520 --> 00:19:00,600
Speaker 2: Really see this issue as one that an packs all

367
00:19:00,640 --> 00:19:03,160
Speaker 2: of us, one that impacts our democracy, one that will

368
00:19:03,200 --> 00:19:05,280
Speaker 2: dictate whether or not we are able to have an

369
00:19:05,320 --> 00:19:08,480
Speaker 2: equitable media and democratic landscape. And when you talk about

370
00:19:08,480 --> 00:19:11,320
Speaker 2: these issues, and please do talk about them, remember that

371
00:19:11,359 --> 00:19:14,200
Speaker 2: it is not just about tailor or social or any

372
00:19:14,240 --> 00:19:18,280
Speaker 2: one person. It's a systemic failure. If girls across the

373
00:19:18,320 --> 00:19:21,359
Speaker 2: globe are being threatened by deep fake technology, that is

374
00:19:21,440 --> 00:19:24,159
Speaker 2: not an individual problem. It is a systemic problem that

375
00:19:24,200 --> 00:19:27,119
Speaker 2: will take a systemic solution. I would also say to

376
00:19:27,200 --> 00:19:29,560
Speaker 2: make sure that you're following the women and women of

377
00:19:29,560 --> 00:19:32,399
Speaker 2: color who have been advocating for making AI safer for

378
00:19:32,480 --> 00:19:35,800
Speaker 2: a long time, Women like doctor Joe Bolomwini, who he

379
00:19:35,840 --> 00:19:38,360
Speaker 2: talked to on the podcast before, because she's been doing

380
00:19:38,400 --> 00:19:41,080
Speaker 2: a lot of work around challenging some of the existential

381
00:19:41,119 --> 00:19:44,000
Speaker 2: threats that AI poses to marginalize people for a long time.

382
00:19:44,440 --> 00:19:48,000
Speaker 2: Check out her advocacy organization, the Algorithmic Justice League. We'll

383
00:19:48,000 --> 00:19:50,160
Speaker 2: put all the links in the show notes and note

384
00:19:50,160 --> 00:19:53,440
Speaker 2: that there is movement on this because of the advocacy

385
00:19:53,480 --> 00:19:56,879
Speaker 2: of people like doctor Boulomwini. The Biden administration released an

386
00:19:56,880 --> 00:20:00,679
Speaker 2: executive order putting some guardrails on AI last year, and

387
00:20:00,800 --> 00:20:04,119
Speaker 2: lawmakers like Rep. Joe Morell and Rep. Evett Clark have

388
00:20:04,320 --> 00:20:09,000
Speaker 2: introduced legislation criminalizing deep thaks. That is because of advocacy,

389
00:20:09,040 --> 00:20:11,720
Speaker 2: and that is because of people speaking up, people using

390
00:20:11,760 --> 00:20:14,199
Speaker 2: their voices to say this is not okay. So we

391
00:20:14,240 --> 00:20:17,440
Speaker 2: need to educate ourselves, stay checked in, and keep making

392
00:20:17,440 --> 00:20:20,680
Speaker 2: it clear that this is not acceptable. You know, I'm

393
00:20:20,720 --> 00:20:24,520
Speaker 2: happy that lawmakers are presenting something, but that is not

394
00:20:24,680 --> 00:20:28,080
Speaker 2: nearly enough. So let's all stay engaged because this fight

395
00:20:28,520 --> 00:20:35,560
Speaker 2: is all of our fights. Got a story about an

396
00:20:35,600 --> 00:20:37,520
Speaker 2: interesting thing in tech or just want to say hi.

397
00:20:37,920 --> 00:20:40,160
Speaker 2: You can reach us at hello at tegodi dot com.

398
00:20:40,240 --> 00:20:42,640
Speaker 2: You can also find transcripts for today's episode at tengody

399
00:20:42,680 --> 00:20:44,760
Speaker 2: dot com. There Are No Girls on the Internet was

400
00:20:44,760 --> 00:20:47,760
Speaker 2: created by me Brigittad. It's a production of iHeartRadio and

401
00:20:47,840 --> 00:20:51,760
Speaker 2: Unbossed Creative. Jonathan Strickland is our executive producer. Tari Harrison

402
00:20:51,800 --> 00:20:54,600
Speaker 2: is our producer and sound engineer. Michael Almado is our

403
00:20:54,600 --> 00:20:58,280
Speaker 2: contributing producer. I'm your host, Brigitad. If you want to

404
00:20:58,280 --> 00:21:00,600
Speaker 2: help us grow, brate and review us on Apple Podcasts.

405
00:21:01,440 --> 00:21:04,240
Speaker 2: For more podcasts from iHeartRadio, check out the iHeartRadio app,

406
00:21:04,280 --> 00:21:06,320
Speaker 2: Apple Podcasts, or wherever you get your podcasts.