Grok Is Still Hosting Sexualized Deepfakes of Famous Women
Grok Is Still Hosting Sexualized Deepfakes of Famous Women
Grok 仍在托管针对知名女性的色情深度伪造内容
Elon Musk’s Grok chatbot is apparently still being used to produce and host nonconsensual explicit images and videos of women, months after Musk’s artificial intelligence firm xAI said it would introduce restrictions to stop the creation of potentially harmful sexualized deepfakes. The revelations come as SpaceX, xAI’s parent company, prepares to go public on Friday in one of the largest IPOs of all time.
埃隆·马斯克（Elon Musk）旗下的聊天机器人 Grok 显然仍在被用于制作和托管未经同意的女性露骨图像和视频。此前，马斯克的人工智能公司 xAI 曾表示将引入限制措施，以阻止制作具有潜在危害的色情深度伪造内容，但几个月过去了，此类情况依然存在。在这些内幕曝光之际，xAI 的母公司 SpaceX 正准备于周五上市，这将是有史以来规模最大的首次公开募股（IPO）之一。
The Grok Imagine generative AI system has been used to create and host images and videos depicting celebrities and at least one politician being held against their will by a giant man, portraying women performing sex acts, and allowing full nudity, a WIRED analysis of public creations found. While some of the images and videos are fully AI-generated or in animated styles, others are photorealistic and show plausible real-world scenarios.
《连线》（WIRED）对公开生成内容的分析发现，Grok Imagine 生成式 AI 系统已被用于创建和托管图像及视频，内容包括描绘名人和至少一位政治人物被一名巨人违背意愿地控制、描绘女性进行性行为，以及展示全裸画面。虽然其中一些图像和视频是完全由 AI 生成或采用动画风格，但另一些则具有照片级的真实感，呈现出看似真实的场景。
WIRED reviewed hundreds of public Grok Imagine links hosted on Grok.com and found dozens led to sexualized AI images and videos, including those created without the subject’s consent. Some links created on Grok.com were subsequently shared on X, including in recent days. The posts, which do not show time stamps of when they were created, are likely just a snapshot of what is being created by people using the Grok Imagine system, as generations do not appear to be made public by default.
《连线》审查了托管在 Grok.com 上的数百个公开 Grok Imagine 链接，发现其中数十个链接指向了色情 AI 图像和视频，包括那些未经当事人同意而创建的内容。一些在 Grok.com 上创建的链接随后被分享到了 X 平台上，甚至在最近几天仍有出现。这些帖子没有显示创建时间戳，可能只是人们使用 Grok Imagine 系统所创建内容的一小部分缩影，因为生成的内容似乎默认情况下并不会公开。
Other generative AI systems deploy more safety guardrails than Grok, which is available on its own website and on X. “While Grok and X may have made some amendments to their model, particularly following the backlash around nudification at the beginning of the year, they still have not done a sufficient job to bring it up to the standard of the other mainstream tools that are available,” says Henry Ajder, an expert on deepfakes who has tracked explicit AI content online for the best part of a decade.
与其他生成式 AI 系统相比，Grok（可在其官网及 X 上使用）部署的安全护栏较少。深度伪造专家亨利·阿杰德（Henry Ajder）表示：“虽然 Grok 和 X 可能对其模型进行了一些修订，特别是在年初因‘去衣化’（nudification）引发强烈抵制之后，但它们在达到其他主流工具标准方面仍做得不够。”阿杰德追踪在线色情 AI 内容已有近十年之久。
Musk’s xAI has faced a wave of lawsuits and scrutiny from regulators around the world since January, when the Grok implementation on X was used to create a flood of “nudification” images. People on X—primarily men—asked the chatbot to edit images to show women in “bikinis” or “string bikinis.” In some instances, images of apparent minors were allegedly also sexualized, according to a class-action lawsuit filed in California federal court in March.
自今年 1 月以来，马斯克的 xAI 面临着全球监管机构的一波诉讼和审查，当时 X 平台上的 Grok 被用于大量生成“去衣化”图像。X 平台上的用户（主要是男性）要求聊天机器人编辑图像，将女性显示为穿着“比基尼”或“丁字比基尼”。根据 3 月份在加州联邦法院提起的一项集体诉讼，在某些情况下，未成年人的图像据称也被色情化了。
Since then, xAI has said it has introduced safeguards to limit and prevent the creation of nonconsensual and sexualized deepfakes. The company has consistently said child sexual abuse material (CSAM) is banned on its platform.
此后，xAI 表示已引入保障措施，以限制和防止制作未经同意的色情深度伪造内容。该公司一贯声称，其平台禁止儿童性虐待材料（CSAM）。
Neither xAI nor X immediately responded to WIRED’s request for comment on the raft of sexualized images hosted on Grok.com. After WIRED contacted the companies, the explicit images and videos WIRED found hosted on Grok.com appeared as no longer being available, and Grok Imagine links shared on X were removed for policy violations. “We strictly prohibit users from generating nonconsensual explicit deepfakes and from using our tools to undress real people,” the safety account on X said in April in response to an NBC report about sexual deepfakes still being created.
xAI 和 X 均未立即回应《连线》关于 Grok.com 上托管大量色情图像的置评请求。在《连线》联系这些公司后，在 Grok.com 上发现的露骨图像和视频似乎已无法访问，而在 X 上分享的 Grok Imagine 链接也因违反政策被删除。X 平台的安全账号在 4 月份回应 NBC 关于色情深度伪造内容持续生成的报道时表示：“我们严禁用户生成未经同意的露骨深度伪造内容，并严禁使用我们的工具对真人进行‘去衣’处理。”
Among the posts hosted on Grok.com, WIRED identified images and videos depicting multiple celebrities as well as US representative Alexandria Ocasio-Cortez. (Spokespeople for Ocasio-Cortez did not immediately respond to WIRED’s request for comment). In the videos, the women were depicted wearing little clothing while being held in the fist of a man’s “giant” hand. One prompt describes a celebrity being held against her will as she “pleads for him not to do this.” It adds: “The giant hand’s grip on her tightens, holding her in place as the giant man holding her leans in and licks her face up and down.”
在 Grok.com 托管的帖子中，《连线》识别出了描绘多位名人以及美国众议员亚历山德里娅·奥卡西奥-科尔特斯（Alexandria Ocasio-Cortez）的图像和视频。（奥卡西奥-科尔特斯的发言人未立即回应《连线》的置评请求）。在视频中，这些女性被描绘成衣着暴露，被一只男性的“巨手”紧紧攥住。其中一个提示词描述了一位名人被违背意愿地控制，她“恳求他不要这样做”。提示词还补充道：“巨手紧紧抓着她，将她固定住，而握住她的巨人俯身，上下舔舐她的脸庞。”
Two prompts that were used to generate material on Grok were rejected by OpenAI’s ChatGPT, Meta AI, and Anthropic’s Claude as inappropriate when tested by WIRED. Google’s Gemini did create an image of one celebrity being held in the hand of a giant, though it rejected another prompt. Google declined to comment.
《连线》测试发现，用于在 Grok 上生成内容的两个提示词，被 OpenAI 的 ChatGPT、Meta AI 和 Anthropic 的 Claude 拒绝，理由是不当。谷歌的 Gemini 虽然拒绝了另一个提示词，但确实生成了一张某位名人被巨人手握住的图像。谷歌拒绝置评。
One Grok Imagine video, which was also posted to X, appeared to depict Ashley St. Clair altered to be dancing in a bikini. St. Clair was previously in a relationship with Musk and is mother to one of his children. In January, she started legal action against xAI after sexualized deepfakes of her allegedly appeared on X. After WIRED contacted X, the post was removed from the social media platform for violating its rules. Legal representatives for St. Clair in the X case did not immediately respond to the request for comment.
一段同样发布在 X 上的 Grok Imagine 视频似乎描绘了阿什莉·圣克莱尔（Ashley St. Clair）身穿比基尼跳舞的画面。圣克莱尔曾与马斯克有过一段恋情，并育有一子。今年 1 月，在 X 上出现关于她的色情深度伪造内容后，她对 xAI 提起了法律诉讼。在《连线》联系 X 平台后，该帖子因违反规则被从社交媒体平台上删除。圣克莱尔在 X 案件中的法律代表未立即回应置评请求。
“Elon Musk knowingly added a perverse feature to his platform that helps users undress women and children at the click of a button, with no regard for the disregard for the predictable damage it would cause,” claims Imran Ahmed, the CEO and founder of the Center for Countering Digital Hate, which in January estimated with a high degree of confidence that Grok created 3 million sexualized images, allegedly including more than 20,000 of children. “Now it appears explicit content is still being hosted on Grok and shared on X, including images ridiculing the mother of Musk’s child.”
“埃隆·马斯克明知故犯地在他的平台上增加了一项变态功能，帮助用户只需点击一下按钮就能让女性和儿童‘脱衣’，完全不顾及这会造成可预见的伤害，”数字仇恨反击中心（Center for Countering Digital Hate）首席执行官兼创始人伊姆兰·艾哈迈德（Imran Ahmed）声称。该中心在 1 月份以极高的置信度估计，Grok 创建了 300 万张色情图像，据称其中包括超过 2 万张儿童色情图像。“现在看来，露骨内容仍在 Grok 上托管并分享到 X 上，甚至包括嘲弄马斯克孩子母亲的图像。”
Unlike other generative AI systems from OpenAI and Google, Musk’s Grok and xAI have not backed away from allowing sexual content in general, having previously introduced “Spicy” and “Unhinged” modes and initially included fewer safety guardrails. Musk has stated Grok is “supposed [to] allow upper body nudity of imaginary adult humans” and be consistent with what viewers might see in R-rated movies. The most recent terms of service from xAI say the system may respond with “sexual situations.” However, the company’s documentation says it does not allow people to use its systems for “causing harm or engaging in abusive activity.”
与 OpenAI 和谷歌的其他生成式 AI 系统不同，马斯克的 Grok 和 xAI 并未完全禁止色情内容。此前，它们曾引入“Spicy”（火辣）和“Unhinged”（狂野）模式，且最初包含的安全护栏较少。马斯克曾表示，Grok “应该允许虚构成年人类的上半身裸露”，并与观众在 R 级电影中可能看到的内容保持一致。xAI 最新的服务条款称，系统可能会回应“性相关情境”。然而，该公司的文档也指出，不允许人们利用其系统“造成伤害或从事虐待活动”。
Other Grok Imagine videos seen by WIRED show women, which are likely entirely AI generated, undressing or involved in sexual acts—some being entirely explicit. The user prompts for many of the videos do not necessarily directly
《连线》看到的其他 Grok Imagine 视频显示，女性（很可能是完全由 AI 生成的）正在脱衣或参与性行为——其中一些内容完全露骨。许多视频的用户提示词并不一定直接指向这些内容。