1 00:00:06,320 --> 00:00:10,160 Just a warning before this episode starts, there will be 2 00:00:10,160 --> 00:00:15,120 some adult type content, some mature type content in this 3 00:00:15,120 --> 00:00:18,720 episode. So don't play it out loud. 4 00:00:19,400 --> 00:00:23,160 For what that's worth. That's about XAI Ashley St. 5 00:00:23,160 --> 00:00:29,000 Clair sexualized images of her swastikas Nazi. 6 00:00:29,720 --> 00:00:35,400 It's it's a lot going on. So that's just your warning and 7 00:00:35,480 --> 00:00:38,520 let's get into it though, because it's interesting to see 8 00:00:38,520 --> 00:00:44,080 where AI is headed as far as image manipulation goes. 9 00:00:44,960 --> 00:00:49,320 So Ashley St. Clair has sued XAI alleging Grok 10 00:00:49,360 --> 00:00:51,960 Image Tool produced sexualized images of her. 11 00:00:52,760 --> 00:00:57,600 Now this case is on Grok's behavior on X and whether XAI 12 00:00:57,600 --> 00:01:01,240 can be held responsible for what its product outputs. 13 00:01:02,000 --> 00:01:04,840 Now, the filing says the AI produced images of Saint Clair 14 00:01:04,840 --> 00:01:07,160 in a string bikini covered with swastikas. 15 00:01:07,520 --> 00:01:10,680 The event depicted her as a child in a bikini. 16 00:01:11,760 --> 00:01:18,360 And is XAI legally responsible for what users make Grok 17 00:01:18,880 --> 00:01:22,000 generate? There's also a number you should 18 00:01:22,000 --> 00:01:24,400 hear. A review said Grok turned out 19 00:01:24,400 --> 00:01:27,680 roughly one non consensual image per minute. 20 00:01:28,760 --> 00:01:31,480 And today we're going to go through what Saint Clair filed, 21 00:01:31,880 --> 00:01:36,720 how X and XAI responded, which governments moved against Grok, 22 00:01:37,080 --> 00:01:41,600 and what could happen next for X and XAI and Grok. 23 00:01:41,600 --> 00:01:45,320 And then we're going to talk about where the pressure points 24 00:01:45,440 --> 00:01:48,840 are still for the platform and for the law. 25 00:01:48,840 --> 00:01:52,280 And we'll go right into that right after this very short 26 00:01:52,280 --> 00:01:55,560 break. Now, Ashley St. 27 00:01:55,560 --> 00:02:00,440 Clair in this lawsuit targets XAI over images Grok generated 28 00:02:00,520 --> 00:02:03,000 of her. She's the mother of one of Elon 29 00:02:03,000 --> 00:02:06,600 Musk's children, and she says the product enabled harassment 30 00:02:06,880 --> 00:02:12,560 that spilled across all of X. The filing tries to make the 31 00:02:12,560 --> 00:02:16,800 maker of the EI directly liable for what the users prompted it. 32 00:02:16,800 --> 00:02:21,440 To do this, the lawsuit describes explicit images, 33 00:02:21,440 --> 00:02:24,240 including one that put her in a string bikini stamped with 34 00:02:24,240 --> 00:02:27,720 swastikas and another that portrayed her as a child in a 35 00:02:27,720 --> 00:02:30,320 bikini. Now that changes the stakes 36 00:02:30,320 --> 00:02:33,640 because the images are both sexualized and, in her account, 37 00:02:34,040 --> 00:02:38,480 weaponized with hateful symbols. The filing also notes the Saint 38 00:02:38,480 --> 00:02:41,200 Clair's Jewish, which heightens the harm described. 39 00:02:42,160 --> 00:02:48,680 Now think of the feed on X and this photo keeps surfacing as 40 00:02:48,680 --> 00:02:52,160 you scrolled through hundreds of these photos were made. 41 00:02:52,560 --> 00:02:57,240 There were different iterations from different people. 42 00:02:57,240 --> 00:03:01,360 They took the original image and tweaked it and also use original 43 00:03:01,360 --> 00:03:03,760 images of her to make these images. 44 00:03:04,520 --> 00:03:07,880 Now her lawyer argues the product itself isn't reasonably 45 00:03:07,880 --> 00:03:10,640 safe and says Grok enabled the abuse by creating and 46 00:03:10,640 --> 00:03:14,240 distributing non consensual images that were then published 47 00:03:14,240 --> 00:03:16,760 on X. That raises a core claim. 48 00:03:16,800 --> 00:03:20,880 In product liability terms, the tool as designed allowed 49 00:03:20,880 --> 00:03:25,360 foreseeable harm. Now if a court up sepsis 50 00:03:25,360 --> 00:03:30,280 framing, you get a direct path to damages for injuries tied to 51 00:03:30,280 --> 00:03:34,040 how the model behaves for Ashley Sinclair. 52 00:03:34,960 --> 00:03:38,360 Now let's just talk about deepfakes here for a second 53 00:03:38,360 --> 00:03:40,080 because we have to understand this term. 54 00:03:40,760 --> 00:03:45,080 It means a synthetic image or video generated by AI that 55 00:03:45,080 --> 00:03:50,520 convincingly looks real. The simple definition matters 56 00:03:50,520 --> 00:03:55,360 because the harm that Sinclair alleges is tied to images that 57 00:03:55,360 --> 00:03:59,840 appear real to just normal people to casual viewers 58 00:04:00,160 --> 00:04:03,880 flipping through X. The filing says Sinclair asked 59 00:04:03,880 --> 00:04:06,080 for removal of the images and filed reports. 60 00:04:06,560 --> 00:04:09,640 She says she then received an e-mail stating there was no 61 00:04:09,640 --> 00:04:12,680 violation in that much of the content stayed public on Grok's 62 00:04:12,680 --> 00:04:16,640 account for over a week. That puts pressure on X's safety 63 00:04:16,640 --> 00:04:20,240 systems because the claim isn't just about that the comment or 64 00:04:20,240 --> 00:04:23,160 that the content just appeared, but that it remained up and they 65 00:04:23,160 --> 00:04:26,600 did nothing about it. The lawsuit also says X's 66 00:04:26,600 --> 00:04:29,600 reporting structure is defective and calls Grok unreasonably 67 00:04:29,600 --> 00:04:35,400 dangerous as designed. That's adds a design and process 68 00:04:35,400 --> 00:04:37,600 claim to the product liability theory. 69 00:04:38,160 --> 00:04:42,000 And if this is all true, that argument makes future incidents 70 00:04:42,000 --> 00:04:45,640 more likely, not less likely, which increases exposure. 71 00:04:46,880 --> 00:04:50,800 And we're going to talk about some definition about guardrails 72 00:04:50,800 --> 00:04:53,040 here. These are built in rules that 73 00:04:53,040 --> 00:04:57,040 try to stop an AI from producing prohibited content. 74 00:04:57,960 --> 00:05:01,960 X, through a statement shared on the platform, said that it 75 00:05:01,960 --> 00:05:05,080 implemented a technical measure to stop the Grok account from 76 00:05:05,080 --> 00:05:08,200 editing images of real people in revealing clothing, like 77 00:05:08,200 --> 00:05:10,240 bikinis. Company also said the 78 00:05:10,240 --> 00:05:13,800 restriction applies to everyone, including paid subscribers. 79 00:05:14,360 --> 00:05:16,880 That limits some obvious prompts, but it doesn't answer 80 00:05:16,880 --> 00:05:19,800 broader questions about how new workarounds get blocked. 81 00:05:20,560 --> 00:05:23,600 So what do we know right now about this? 82 00:05:24,080 --> 00:05:27,360 I did an experiment last night because I was researching a 83 00:05:27,360 --> 00:05:32,680 story and I was just wondering what would happen if I typed 84 00:05:32,800 --> 00:05:40,360 into X, typed into Grok, and I said make an image of Donald 85 00:05:40,360 --> 00:05:45,600 Trump in a bikini. Sure enough, 3 seconds later, 86 00:05:46,560 --> 00:05:51,120 Donald Trump, the President of the United States, was in a very 87 00:05:51,120 --> 00:05:55,600 revealing bikini. And I thought that's horrible. 88 00:05:55,840 --> 00:05:57,920 That is disgusting. Actually, I didn't want to see 89 00:05:57,920 --> 00:06:00,640 that. And then I thought, well, maybe 90 00:06:00,640 --> 00:06:05,240 this guardrails up because Elon Musk doesn't want people doing 91 00:06:05,240 --> 00:06:07,960 that to him. So I did it for, of course, 92 00:06:07,960 --> 00:06:11,200 research sake. I typed the same thing into 93 00:06:11,200 --> 00:06:15,000 Grok. I said make, you know, make a 94 00:06:15,120 --> 00:06:18,880 image of Elon Musk in a bikini. Sure enough, a few seconds 95 00:06:18,880 --> 00:06:23,760 later, there was an image of Elon Musk in a bikini and his 96 00:06:23,760 --> 00:06:25,920 body had morphed and he's filled out the bikini. 97 00:06:25,920 --> 00:06:28,840 Same thing with Donald Trump. And it was disgusting. 98 00:06:29,200 --> 00:06:32,160 I just want to see if these guardrails guardrails were 99 00:06:32,160 --> 00:06:38,400 actually up and they are not. These revealing images are still 100 00:06:38,480 --> 00:06:41,360 allowed on X. So Ashley St. 101 00:06:41,360 --> 00:06:45,360 Clair has a point. There are no guardrails even 102 00:06:45,360 --> 00:06:48,320 now. And this is after X has actually 103 00:06:48,520 --> 00:06:52,520 released a statement saying that this is not possible anymore. 104 00:06:53,720 --> 00:06:59,120 So XS safety team instead of removes child abuse material and 105 00:06:59,120 --> 00:07:01,600 non consensual nudity and reports offenders to law 106 00:07:01,600 --> 00:07:04,720 enforcement when necessary. It opens a compliance lane for 107 00:07:04,720 --> 00:07:07,200 the platform and if the rules are clear and enforced, the 108 00:07:07,200 --> 00:07:10,200 company can point to an enforcement regime. 109 00:07:10,200 --> 00:07:14,520 So the lawsuit, however, challenges both the speed and 110 00:07:14,520 --> 00:07:18,640 effectiveness of these actions. In this case, it seems like 111 00:07:18,640 --> 00:07:24,240 there are not many guardrails on X right now, but if there are 112 00:07:24,440 --> 00:07:27,840 and they want to enforce them, they have the ability to do so. 113 00:07:28,720 --> 00:07:32,520 So they see Sam that we have to talk about C Sam too, because 114 00:07:32,800 --> 00:07:34,360 some people aren't familiar with this. 115 00:07:34,560 --> 00:07:38,160 It's an acronym for Child Sexual Abuse Material, which is illegal 116 00:07:38,160 --> 00:07:41,720 content involving minors. Platforms that encounter CSM 117 00:07:41,760 --> 00:07:44,560 must remove it and report it to authorities, which adds 118 00:07:44,560 --> 00:07:47,040 substantial legal risk if they fail to do so. 119 00:07:47,840 --> 00:07:51,760 Now, does that count with somebody making a prompt? 120 00:07:52,480 --> 00:07:54,040 That's what regulators are going to work on. 121 00:07:54,440 --> 00:07:58,040 Grok has reportedly blocked in Malaysia and Indonesia, and 122 00:07:58,040 --> 00:08:01,600 authorities in the UK and in California, where the company is 123 00:08:01,600 --> 00:08:06,560 based, are investigating the company that changes the 124 00:08:06,560 --> 00:08:09,240 operational map. Because a blocked market in 125 00:08:09,280 --> 00:08:14,560 active pros can force product changes, can slow down releases 126 00:08:14,600 --> 00:08:20,480 or add more legal obligations to the company, the venue is no 127 00:08:20,480 --> 00:08:24,240 longer just one U.S. court. The content analysis firms 128 00:08:24,240 --> 00:08:28,400 review estimated that Grok was generating roughly one non 129 00:08:28,400 --> 00:08:32,039 consensual image per minute during a recent spike. 130 00:08:33,120 --> 00:08:36,400 That is it at scale, which is exactly what worries regulators, 131 00:08:36,400 --> 00:08:40,320 plaintiffs and advertisers. If that number holds, it makes 132 00:08:40,320 --> 00:08:42,520 the harm systemic rather than anecdotal. 133 00:08:42,520 --> 00:08:44,880 It's not just one person, it's the whole platform. 134 00:08:45,120 --> 00:08:47,120 It's just scummy people all over the platform. 135 00:08:47,120 --> 00:08:51,200 You have a lawsuit, ongoing investigations, and now you have 136 00:08:51,200 --> 00:08:54,240 a volume problem. You have millions of these 137 00:08:54,240 --> 00:08:57,120 images out there all over the platform. 138 00:08:57,120 --> 00:09:00,880 How does Elon Musk and his team battle this? 139 00:09:01,800 --> 00:09:05,120 Elon Musk posted that he was not aware of naked underage images 140 00:09:05,120 --> 00:09:08,520 generated by Grok and said the system only produces images when 141 00:09:08,520 --> 00:09:12,520 users ask it to. He added the Grok refuses 142 00:09:12,520 --> 00:09:15,920 illegal requests and that adversarial prompt hacking can 143 00:09:15,920 --> 00:09:18,160 trigger unexpected behavior. That engineers. 144 00:09:18,360 --> 00:09:24,480 Then we'll patch that defense is not that good to be honest with 145 00:09:24,480 --> 00:09:26,000 you. You should have these guardrails 146 00:09:26,000 --> 00:09:30,640 up way before anybody does anything horrible like this. 147 00:09:31,560 --> 00:09:37,480 Any Elon Musk is saying that Grok as a tool obeys laws but 148 00:09:37,480 --> 00:09:42,600 sometimes encounters some edge cases and therefore they can't 149 00:09:42,600 --> 00:09:44,520 really do anything about it, right. 150 00:09:45,680 --> 00:09:49,840 The filing says that Sinclair sought both compensary and 151 00:09:49,840 --> 00:09:53,560 punitive damages. Compensatory damages repay 152 00:09:53,560 --> 00:09:56,840 losses like emotional distress or lost income, or punitive 153 00:09:56,840 --> 00:09:59,440 damages are meant to punish and deter. 154 00:10:00,200 --> 00:10:04,360 That opens a big financial risk if a court decides the conduct 155 00:10:04,360 --> 00:10:06,480 warrants punishment beyond compensation. 156 00:10:07,360 --> 00:10:10,680 Punitive damages are punishment money, not reimbursement money. 157 00:10:11,280 --> 00:10:14,400 Now, the lawsuit also alleges some X actions that go beyond 158 00:10:14,400 --> 00:10:18,160 removal decisions and says her account was demonetized and that 159 00:10:18,160 --> 00:10:21,520 more images of her were generated after she objected. 160 00:10:22,560 --> 00:10:25,160 Now, if that's accurate, the strength is the claim that the 161 00:10:25,160 --> 00:10:29,160 system retaliated and that the company hadn't noticed an 162 00:10:29,280 --> 00:10:34,440 ongoing system of harm. So these claims cover design 163 00:10:34,480 --> 00:10:37,600 moderation and alleged retaliation. 164 00:10:37,600 --> 00:10:42,160 Now, and I've been digging through the analytics of this 165 00:10:42,160 --> 00:10:46,080 show right now, and I've noticed that 37% of you are following 166 00:10:46,080 --> 00:10:48,600 the show. And for you, I'm your biggest 167 00:10:48,600 --> 00:10:50,480 fan. Thank you so much for following 168 00:10:50,480 --> 00:10:52,000 the show, hitting the subscribe button. 169 00:10:52,480 --> 00:10:54,880 Now there's 63% of you that probably listen to the show 170 00:10:54,880 --> 00:10:57,640 sometimes and you might have listened to two or three 171 00:10:57,640 --> 00:11:00,720 episodes and just forgot to hit the subscribe or follow button. 172 00:11:01,120 --> 00:11:04,000 Or you're new here and you'd like the show and you want more 173 00:11:04,000 --> 00:11:07,120 like this. You want more info about SpaceX, 174 00:11:07,120 --> 00:11:12,520 NASA, spaceflight, Tesla, Elon XAI, Grok, all the stuff, right? 175 00:11:13,040 --> 00:11:15,880 And I've been an independent journalist covering all of that, 176 00:11:16,000 --> 00:11:19,280 Elon Musk, spaceflight, SpaceX, etcetera, for 6 plus years. 177 00:11:19,280 --> 00:11:23,880 I actually embedded myself for my YouTube channel at SpaceX, at 178 00:11:23,880 --> 00:11:27,400 Boca Chica, Texas, at Starbase for about a year on the side of 179 00:11:27,400 --> 00:11:29,240 the road because I love this so much. 180 00:11:29,760 --> 00:11:32,400 So I've been doing this for like 6 years and I intend on 181 00:11:32,520 --> 00:11:34,600 continuing doing it for the next 10. 182 00:11:35,160 --> 00:11:39,880 And all I ask from you is one second of your time to hit the 183 00:11:39,880 --> 00:11:42,840 follow up subscribe button because it helps us so, so much 184 00:11:43,000 --> 00:11:46,400 on this platform. And with your support, we can 185 00:11:46,400 --> 00:11:48,920 continue to grow and do better things. 186 00:11:49,160 --> 00:11:52,680 So if you want to help us do better, please hit the subscribe 187 00:11:52,760 --> 00:11:55,280 or follow button on the platform that you're listening or 188 00:11:55,280 --> 00:11:58,000 watching on right now. I'm extremely grateful and 189 00:11:58,000 --> 00:11:59,800 blessed to have you here in this community. 190 00:11:59,800 --> 00:12:01,480 So thank you, thank you, thank you. 191 00:12:03,360 --> 00:12:05,160 Now let's go back into this filing. 192 00:12:05,480 --> 00:12:08,440 It also asserts that Crock and Xai had explicit knowledge that 193 00:12:08,440 --> 00:12:11,400 she did not consent because she told the company to remove the 194 00:12:11,400 --> 00:12:14,640 images. This is a notice element now, 195 00:12:14,720 --> 00:12:17,800 which in many cases is a threshold for liability if 196 00:12:17,800 --> 00:12:21,360 harmful content stays up. If the company knew and failed 197 00:12:21,360 --> 00:12:25,600 to act quickly, the plaintiffs argument gets stronger. 198 00:12:25,600 --> 00:12:28,320 So Ashley St. Clair's argument gets much 199 00:12:28,520 --> 00:12:32,360 stronger because she asked them to remove it and they did 200 00:12:32,360 --> 00:12:35,160 nothing. So there's other contacts too. 201 00:12:35,280 --> 00:12:39,600 Her family, Musk and Saint Clair have a son born in 2024 and are 202 00:12:39,600 --> 00:12:42,040 estranged. And Musk has signaled he plans 203 00:12:42,040 --> 00:12:45,680 to seek custody now. So that doesn't determine 204 00:12:46,400 --> 00:12:49,640 liability, but it adds public attention and can shape how both 205 00:12:49,640 --> 00:12:52,480 sides talk about the case, including the courts. 206 00:12:52,880 --> 00:12:56,320 Meanwhile, XS public statements keep returning to rules and 207 00:12:56,320 --> 00:12:58,400 enforcement. The company says it removes high 208 00:12:58,400 --> 00:13:01,760 priority violations and zeroes out non consensual sexual 209 00:13:01,760 --> 00:13:04,840 content. And it provides points to 210 00:13:04,840 --> 00:13:10,040 reporting to law enforcement for C Sam dispositions X as 211 00:13:10,040 --> 00:13:12,200 proactive. But the lawsuit argues the 212 00:13:12,200 --> 00:13:14,640 experience on the ground was slower and less effective. 213 00:13:15,400 --> 00:13:18,600 So where does this go next? Though Musk had not publicly 214 00:13:18,600 --> 00:13:22,520 responded to the lawsuit as of today, which suggests an 215 00:13:22,520 --> 00:13:26,600 initial, you know, his replies may appear on the platform 216 00:13:26,600 --> 00:13:29,120 rather than informal court filings, which usually speaks 217 00:13:29,120 --> 00:13:32,440 out publicly before the court. But maybe not this time. 218 00:13:32,800 --> 00:13:39,200 But he has said that X is an XAI and Grok and X have all been 219 00:13:39,200 --> 00:13:41,240 working together to kind of battle this. 220 00:13:41,240 --> 00:13:44,600 But also it's up to the users to generate the content, so they 221 00:13:44,600 --> 00:13:47,840 can't really police it. I guess that's what happens when 222 00:13:47,840 --> 00:13:52,240 you destroy the safety team when you take over Twitter. 223 00:13:52,240 --> 00:13:55,880 The investigation still continues in the UK and 224 00:13:55,880 --> 00:13:58,840 California. And that suggests this issue 225 00:13:58,840 --> 00:14:02,240 will evolve through multiple venues, not just one courtroom, 226 00:14:02,440 --> 00:14:07,200 and overseas as well. So this is a filing that accuses 227 00:14:07,200 --> 00:14:10,440 an AI product of generating sexualized, hateful and in one 228 00:14:10,440 --> 00:14:14,440 case, childlike images of a real person, and it connects those 229 00:14:14,440 --> 00:14:18,680 images to moderation and design choices on a major platform. 230 00:14:19,160 --> 00:14:22,160 The company says it has safety measures, it removed violations, 231 00:14:22,160 --> 00:14:25,560 and it avoids illegal outputs. While the plaintiff says those 232 00:14:25,560 --> 00:14:28,760 steps did not work when it mattered, the next phase will 233 00:14:28,760 --> 00:14:32,360 test whether courts treat AI image engines like ordinary 234 00:14:32,360 --> 00:14:36,480 software or like products with a duty to prevent specific harms 235 00:14:36,760 --> 00:14:37,480 for everybody.