Despite previous assurances to crack down on harmful content, Elon Musk’s Grok chatbot from xAI continues to be implicated in the creation and hosting of nonconsensual explicit images and videos of women. These concerning revelations surface just as SpaceX, the parent company of xAI, prepares for one of the largest IPOs in history.
A recent analysis by WIRED uncovered numerous public Grok Imagine links that led to sexualized AI-generated content, including depictions of celebrities and even politicians in compromising situations. These creations ranged from fully AI-generated or animated styles to shockingly photorealistic scenarios, some portraying individuals performing sex acts or appearing in full nudity.
Persistent Deepfake Concerns
WIRED’s extensive review of hundreds of public Grok.com links identified dozens featuring nonconsensual explicit AI images and videos. While posts shared on X (formerly Twitter) don’t always show timestamps, they offer a snapshot of the potentially vast amount of content being generated by users of the Grok Imagine system, as creations are not public by default.
Experts in the field, like Henry Ajder, who has tracked explicit AI content for a decade, note that Grok’s safety guardrails still fall short compared to other mainstream generative AI tools. Despite some amendments made after initial backlash, Grok, available on its website and X, continues to lag in preventing the creation of such harmful content.
Since January, xAI has faced a barrage of lawsuits and regulatory scrutiny globally, particularly after a wave of “nudification” images emerged on X. These images, often requested by users to depict women in “bikinis” or “string bikinis,” allegedly included sexualized images of apparent minors, leading to a class-action lawsuit filed in California federal court in March.
Although xAI has stated its commitment to introducing safeguards and explicitly banning child sexual abuse material (CSAM), the persistence of nonconsensual content raises serious questions. Following WIRED’s inquiries, many of the explicit images and videos found on Grok.com were made unavailable, and infringing links on X were removed for policy violations. The safety account on X reiterated its stance in April, stating, “We strictly prohibit users from generating nonconsensual explicit deepfakes and from using our tools to undress real people.”
Targeted Individuals and Industry Contrast
Among the unsettling content hosted on Grok.com, WIRED identified images and videos depicting multiple public figures, including US Representative Alexandria Ocasio-Cortez. These deepfakes often showed women in minimal clothing, held captive in the fist of a giant man, with one prompt disturbingly describing a celebrity pleading as a giant hand tightens its grip and licks her face.
A stark contrast emerges when comparing Grok’s content generation capabilities with those of its competitors. Prompts that Grok readily accepted for generating explicit material were rejected as inappropriate by leading AI systems like OpenAI’s ChatGPT, Meta AI, and Anthropic’s Claude. Even Google’s Gemini, while accepting one similar prompt, declined another, highlighting Grok’s more permissive approach.
Another deeply concerning case involved a Grok Imagine video, also posted on X, that appeared to depict Ashley St. Clair, a former partner of Elon Musk and mother to one of his children, dancing in a bikini. St. Clair initiated legal action against xAI in January after sexualized deepfakes of her allegedly appeared on X. This post was subsequently removed after WIRED contacted X.
Critics, such as Imran Ahmed, CEO of the Center for Countering Digital Hate (CCDH), assert that Elon Musk knowingly added a “perverse feature” to his platform. CCDH estimates that Grok generated 3 million sexualized images, including over 20,000 of children, underscoring the severe and widespread impact of these functionalities.
Legal Headwinds and Regulatory Scrutiny
Unlike many other generative AI platforms that have distanced themselves from sexually explicit content, Musk’s Grok and xAI have taken a different path. They previously introduced “Spicy” and “Unhinged” modes and initially offered fewer safety guardrails. Musk himself has stated that Grok is “supposed [to] allow upper body nudity of imaginary adult humans” and be consistent with R-rated movie content.
While xAI’s terms of service acknowledge that the system may respond with “sexual situations,” the company’s documentation also prohibits use for “causing harm or engaging in abusive activity.” Researchers observe that some changes since January have made it harder to create direct “nudification” images of real people, with a noticeable decrease in such content on X.
However, the ongoing risks are significant enough that SpaceX, ahead of its IPO, has set aside $530 million to address potential legal complaints, including those linked to Grok. The company’s filing acknowledges risks such as “reputational harm, the generation of potentially explicit content and misinformation or deceptive outputs, potential nonconsensual or exploitative imagery.”
Further compounding these challenges, the Privacy Commissioner of Canada recently published preliminary findings from an investigation into xAI and Grok’s deepfake transgressions. The investigation alleges that xAI violated Canada’s federal private-sector privacy law by failing to implement “appropriate safeguards from the outset.” Despite xAI’s claims of introducing new checks and safeguards, the Privacy Commissioner remains unconvinced of their effectiveness in mitigating these critical issues.
Source: Wired – AI
