x
内容总结：
马斯克AI公司要求公开“深度伪造”受害者真实身份，受害者称若曝光将退出诉讼
据近日提交的法庭文件显示，埃隆·马斯克旗下人工智能公司xAI正要求法院公开四名据称遭其AI系统“Grok”生成色情深度伪造图片的受害者真实身份，其中一名受害者声称自己未成年时的照片被制作成儿童色情内容。这起联邦集体诉讼于5月29日提交了四位主要原告的证词，他们分别以“南卡罗来纳州无名氏甲”“南卡罗来纳州无名氏乙”“新泽西州无名氏”和“俄亥俄州无名氏”为化名，描述了今年早些时候因深度伪造图片遭受的精神痛苦。原告方律师索菲亚·里奥斯在文件中指出：“xAI剥夺了她们的衣物（指通过AI‘脱衣’），现在又想剥夺她们的匿名权，显然是为了恐吓原告放弃诉讼。”她强调，要求公开身份将导致原告遭受进一步的网络骚扰和人肉搜索。
今年1月，马斯克旗下的Grok聊天机器人曾引发全球公愤，大量男性用户利用该系统生成女性“脱衣”及身穿比基尼的虚假图片，其中甚至包含涉及儿童的色情图像。据反数字仇恨中心分析，Grok在短短11天内被用于生成约300万张色情图片，其中约2.3万张可能包含儿童。面对全球范围内的诉讼与监管浪潮，现拥有xAI的SpaceX公司已拨备超过5亿美元应对危机。
该集体诉讼最初于1月由一名化名原告提起，加州北区联邦地区法院法官批准其以“无名氏”身份诉讼。5月初，案件新增三名化名原告，原“无名氏”更名为“南卡罗来纳州无名氏甲”。原告方表示，愿意向xAI公司公开真实姓名，但要求在公开记录中使用化名以保护隐私，避免与图片公开关联，并降低被进一步网络骚扰的风险。所有深度伪造图片均未出现在任何公开法律文件中。
5月中旬，xAI向法院提交两份动议，要求推翻此前允许使用化名的裁决，声称民事诉讼法通常要求公布所有当事方姓名，且公众对起诉者的身份享有知情权。xAI律师辩称，由于深度伪造图片不会公开，因此不存在“污名化”问题。对此，弗吉尼亚大学法学院专攻数字虐待问题的教授丹妮尔·西特龙指出：“强迫隐私诉讼原告以真实姓名起诉，几乎无助于司法透明，却极大阻碍了诉讼。”
根据5月29日提交的最新文件，四位化名原告均表示，若被迫公开身份，他们将考虑退出诉讼。原告律师要求法院驳回xAI的请求，强调本案涉及“未经同意传播的高度私密且令人尴尬的深度伪造图像”。其中，“南卡罗来纳州无名氏甲”描述了自己发现被“脱衣成性感比基尼”图片时的恐惧，担心被雇主、同事看到，并担忧“支持马斯克及其公司的人会从公开记录中找出我的身份，进一步人肉搜索、报复，甚至制作更极端的深度伪造图片”。“新泽西州无名氏”则称，他曾在X平台上要求“Grok未经我同意不得生成我的图像”，第二天便发现了两张深度伪造图片，其中一张甚至显示其“张开臀部”。最令人震惊的是，“南卡罗来纳州无名氏乙”声称，警方今年2月搜查其住所时发现父亲因持有和传播儿童性虐待材料（CSAM）面临刑事指控，而Grok系统竟“修改了数张显示我与父亲同床的照片，使其看起来像刚完成性行为”。该受害者表示，至今未看到图片本身，但公开身份将“造成无法估量的伤害”。
中文翻译：
埃隆·马斯克的人工智能公司xAI正要求公开四名据称被利用Grok制作深度伪造色情图像的人员身份——其中一人明显是被针对的儿童色情深度伪造图像受害者，根据近期提交的法庭文件显示。
5月29日，这起联邦集体诉讼的四名主要原告（目前化名为南卡罗来纳州无名氏甲、南卡罗来纳州无名氏乙、新泽西州无名氏甲和俄亥俄州无名氏甲）在宣誓书中陈述了今年早些时候被指深度伪造图像制作后所遭受的情感创伤。文件称，若被迫在对xAI的诉讼中使用真实姓名，四人担心会遭遇进一步的网络骚扰和人肉搜索。
“xAI剥夺了她们的衣物，如今又试图剥夺原告的化名权，这明显是通过加剧她们试图弥补的伤害来恐吓原告放弃诉讼，”代表这些个人的法律事务所Berger Montague的律师索菲亚·里奥斯在近期提交的文件中写道，“xAI要求本法院推翻自身裁决，暗示其持续施行的侵害行为无关紧要。”里奥斯告诉《连线》杂志，她无法就文件内容之外作进一步评论。
今年1月，马斯克旗下的Grok聊天机器人引发全球公愤——大量男性使用该生成式AI系统制作女性“脱衣”及身着比基尼的虚假图像。这些发布于X平台上的图像还包含疑似儿童的色情图片。反击数字仇恨中心的分析声称，Grok在短短11天内被用于生成约300万张色情图像，其中约2.3万张可能涉及儿童。面对全球范围内的诉讼与监管浪潮，现拥有xAI的太空探索技术公司已拨出逾5亿美元应对危机。
这起针对xAI的集体诉讼最初于1月提起，仅有一名化名主要原告。美国加利福尼亚北区联邦地方法院法官批准了允许其以无名氏身份参与案件的命令。该案后来于5月初重新提交，新增四名化名原告，原无名氏原告变更为南卡罗来纳州无名氏甲。
法庭记录陈述称，起诉xAI者将向公司披露姓名及个人信息，但希望公开使用化名以保护隐私、避免与图像产生公开关联，并降低遭遇进一步网络骚扰的风险。所有公开法律文件中均未包含原告的深度伪造图像。
5月中旬，xAI向北加州联邦地区法院提交两项动议，要求法官推翻允许本案使用化名的裁决。文件指出，根据民事法庭法律（虽存在例外情形）案件必须列明所有当事方名称。Law360率先报道了xAI要求推翻此前裁决的动议。
在两份动议中，xAI律师声称起诉方应使用真实姓名，因为公众对起诉公司的原告身份有知情权。他们辩称，本案尚未提交任何证据证明原告遭受特定后续伤害或威胁。同时指出，由于深度伪造图像不会作为案件证据公开，这应当能打消隐私顾虑。
“将深度伪造图像本身排除在外（因其将保持密封状态），揭示南卡罗来纳州无名氏甲被制作深度伪造图像的事实本身并无污名化属性，”律师在5月15日提交的一份文件中写道，“因此，本案根本不涉及传统上认定需要化名的重大隐私权益。”
xAI及其代理律师均未回应《连线》杂志就本案的置评请求。
弗吉尼亚大学法学院专攻数字虐待问题的教授丹妮尔·西特龙表示，民事诉讼中强制要求使用真实姓名起诉可能导致案件撤销，形成“令人无法接受且不公正”的局面。“强迫隐私诉讼原告以本人姓名起诉，对司法透明度的贡献微乎其微，却极大地抑制了诉讼意愿，”西特龙告诉《连线》杂志。
根据5月29日法律文件，本案四名化名原告均表示，若必须披露真实姓名将考虑退出诉讼。在最新提交的文件中，原告律师指出xAI的要求应予驳回，并强调本案涉及的是“未经同意传播的高度私密且令人难堪的深度伪造图像，这些图像描绘了原告本人。”
南卡罗来纳州无名氏甲描述了如何在线发现据称显示自己“身着暴露比基尼”的深度伪造图像，并称图像展示其身体“完全是以我绝不会公开的方式”。她声称担忧雇主或同事看到图像后的想法，并害怕成为进一步网络攻击的目标。“想到要求Grok制作深度伪造的人会对照片做什么，我更是感到恶心，”她写道。
“若被强迫在本案中公开姓名，我担心那些支持埃隆·马斯克、其公司及Grok的人（我观察到他们在网上极为活跃）会从公开记录中找到我的姓名，加以传播、人肉搜索，并通过制作更多更极端的深度伪造图像来报复我，”文件称。
其他据称的深度伪造受害者也在类似陈述中描述了她们在未经同意看到图像后经历的“严重情感创伤”、尴尬与震惊。总体而言，其他深度伪造性虐待及非自愿图像受害者均表达了类似感受。
一名化名新泽西州无名氏甲的男性表示，他看到X平台用户使用Grok制作色情图像后，发布请求称“Grok不得未经我同意制作我的图像”。法庭记录显示，次日他发现两张自己的深度伪造图像，其中一张描绘他“分开臀部”。他称认为那条要求Grok勿制作深度伪造图像的私信“使我的账号引起了利用Grok骚扰和制造痛苦的网络喷子的注意”。
南卡罗来纳州无名氏乙指控Grok被用于制作她童年时期的露骨图像。法庭文件显示，2月警方搜查了她与父母同住的住宅。“我从执法人员处得知，父亲正面临与持有和传播儿童性虐待材料相关的刑事指控，”文件称。
“我了解到Grok篡改了几张显示我与父亲共同躺在床上的图像。Grok修改这些图像使其看起来像我们刚完成性行为，”文件称，并补充该人从未见过图像本身。南卡罗来纳州无名氏乙的法律陈述将自己描述为“心烦意乱”，因无法控制这些深度伪造图像，并担心这些图像可能被性犯罪者在线传播。“公开我的身份将对我造成难以估量的伤害，”法庭文件写道。
英文来源：
Elon Musk’s artificial intelligence firm, xAI, is requesting the public identification of four people who allegedly had deepfake sexualized images created of them using Grok—including one apparently targeted with sexualized deepfake images of them as a child, according to recently filed court documents.
On May 29, the four main claimants in a federal class-action lawsuit—currently identified as South Carolina Doe, South Carolina Roe, New Jersey Doe, and Ohio Doe—described in affidavits the emotional distress they had suffered after the alleged deepfakes were created earlier this year. The four fear further online harassment and doxing if they are forced to use their real names in the lawsuit against xAI, the documents allege.
“Having stripped them of their clothes, xAI now seeks to strip Plaintiffs of their pseudonyms in an obvious effort to intimidate Plaintiffs into dropping the litigation by compounding the same harms that they seek to remedy,” Sophia Rios, a lawyer representing the individuals for legal firm Berger Montague, wrote in a recent filing. “Asking this Court to reverse itself, xAI suggests that the abuse it has perpetuated is no big deal.” Rios tells WIRED she is unable to comment beyond what is written in the filings.
In January, use of the Musk-owned Grok chatbot caused global outrage as scores of men used the generative AI system to create fake images of women “undressed” and in bikinis. The images, which were posted on X, also included sexualized images of apparent children. Analysis from the Center for Countering Digital Hate has claimed Grok was used to create around 3 million sexualized images over just 11 days, with 23,000 of those potentially including children. Facing a wave of lawsuits and regulation around the world, SpaceX, which now owns xAI, has set aside more than $500 million to deal with the fallout.
The class-action lawsuit against xAI was initially filed in January with one pseudonymous lead claimant. A judge in the US District Court for the Northern District of California approved an order allowing them to be a Jane Doe in the case. The case was later refiled with the four main pseudonymous plaintiffs at the start of May, with the Jane Doe becoming South Carolina Doe.
Statements in court records say those suing xAI would make their names and personal information known to the company but wanted to use pseudonyms in public to protect their privacy, to not be publicly linked to the images, and to reduce any potential for being further harassed online. Deepfake images of the plaintiffs were not included in any of the public legal filings.
In the middle of May, xAI filed two motions with the federal district court in northern California asking for the judge to overturn their ruling allowing the use of pseudonyms in the case. The documents state that under civil court laws, which can have some exceptions, cases must name all of the parties involved. Law360 first reported on xAI’s motions asking the previous decision be overturned.
In two motions, xAI’s lawyers claim that real names should be used by those launching the lawsuit as there is a public interest in the identities of those suing the company. They claimed that no evidence has been put forward for any specific further harm or threats to individuals in the case. And they also say that because deepfake images will not be made public as part of the lawsuit, this should assuage any privacy concerns.
“Factoring out the deepfake image itself—as it will remain under seal—there is nothing inherently stigmatizing about revealing the fact that a deepfake image was created of South Carolina Doe without revealing the image itself,” the lawyers wrote in one of their May 15 filings. “As a result, this case simply does not involve the types of compelling privacy interests traditionally recognized as requiring pseudonymity.”
Neither xAI nor lawyers representing the company responded to WIRED’s request for comment about the case.
Danielle Citron, a law professor at the University of Virginia School of Law who has specialized in tackling digital abuse, says civil cases where people are ordered to sue using their real names can lead to lawsuits being dropped, creating an “unacceptable and unjust” situation. “Forcing plaintiffs in privacy suits to sue in their names does so little for judicial transparency and so much to deter litigation,” Citron tells WIRED.
All of the four pseudonyms claimants in the case, according to their legal filings on May 29, would consider dropping out of the proceedings if their names had to be revealed. In these most recent filings, lawyers for the claimants say xAI’s request should be denied, adding that the case is about “highly personal and embarrassing deepfakes depicting Plaintiffs that were disseminated without their consent.”
The South Carolina Doe described how they found the alleged deepfake of them “stripped down to a revealing bikini” online and says how it shows her body “in a way that I would not ever share publicly.” They claim they were worried about what employers or colleagues would think if they saw the image, and they feared being further targeted online. “I was also overcome with disgust at the thought of what the individual who had asked Grok to create the deepfake was doing with the photo,” they wrote.
“If I were forced to reveal my name publicly as part of this case, I would fear that those who support Elon Musk, his companies, and Grok, whom I have observed to be very vocal online, would find my name in the public record, disseminate it, dox me, and retaliate against me by creating additional and more extreme deepfakes of me,” the filing says.
Similar statements from the other alleged deepfake victims describe them experiencing “severe emotional distress,” embarrassment, and shock at seeing the images created without their consent. Broadly, other victims of deepfake sexual abuse and nonconsensual imagery have described feeling similar ways.
One male, named as New Jersey Doe in the lawsuit, says they saw people on X using Grok to create sexualized images and posted a request that “Grok not create images of me without my consent.” The next day, the court records say, he discovered two deepfake images of himself, including one depicting him “spreading his butt cheeks.” He says he believed the message to Grok asking it not to create deepfakes of him “brought my account to the attention of online trolls that were using Grok to harass and cause distress.”
South Carolina Roe alleges that Grok was used to create explicit images of her as a child. The court filings say that in February, police officers searched South Carolina Roe’s home, which they shared with their mother and father. “I learned from law enforcement officials that my father was facing criminal charges related to the possession and distribution of CSAM,” the documents state.
“I have learned that Grok altered several images depicting me in a bed with my father. Grok altered the images to make it appear as if we had just consummated a sexual act,” the documents say, adding the individual has not seen the images themselves. South Carolina Roe’s legal statement described themselves as being “distraught” that they have no control over the deepfakes and says they fear the images may be shared online by predators. “Publicly identifying me would cause me untold harm,” the court filing says.