Elon Musk, CEO of xAI, on Wednesday denied awareness of any naked underage images generated by Grok, his company's AI chatbot. This statement came hours before the California Attorney General, Rob Bonta, launched an investigation into xAI regarding the "proliferation of nonconsensual sexually explicit material" produced by Grok. The investigation and mounting global pressure from governments in the U.K., Europe, Malaysia, and Indonesia follow numerous reports of users prompting Grok to transform photos of real women, and in some instances children, into sexualized images without consent.

Copyleaks, an AI detection and content governance platform, estimated the scale of the issue, reporting roughly one such image posted on X (formerly Twitter, also part of the same company as xAI) each minute. A more focused sample taken from January 5 to January 6 revealed an alarming rate of 6,700 such images per hour over a 24-hour period. Attorney General Bonta emphasized the severity of the problem, stating that this material has been used to harass people online, and urged xAI to take immediate action to prevent further harm. The AG's office will investigate potential violations of laws designed to protect individuals from nonconsensual sexual imagery and child sexual abuse material (CSAM).

Several legal frameworks exist to address these issues. The federal Take It Down Act, signed into law last year, criminalizes the knowing distribution of nonconsensual intimate images, including deepfakes, and mandates platforms like X to remove such content within 48 hours. California also enacted its own series of laws in 2024, signed by Governor Gavin Newsom, specifically targeting sexually explicit deepfakes.

The trend of Grok generating sexualized images of women and children began towards the end of the previous year. It reportedly gained traction after certain adult-content creators used Grok to generate sexualized imagery of themselves for marketing purposes, prompting other users to follow suit. There have been public instances, including well-known figures like actress Millie Bobby Brown, where Grok altered real photos of women by changing clothing, body positioning, or physical features in overtly sexual ways.

Reports suggest that xAI has started implementing safeguards to tackle these problems. Grok now requires a premium subscription for certain image-generation requests, and even then, image generation may not occur. April Kozen, VP of marketing at Copyleaks, noted that Grok might fulfill requests in a more generic or subdued manner, though it appears more permissive with adult content creators. Kozen added that these behaviors indicate X is experimenting with mechanisms to control problematic image generation, despite remaining inconsistencies.

Neither xAI nor Elon Musk has directly addressed the core issue of Grok's role in generating such content. Days after the incidents surfaced, Musk appeared to downplay the situation by asking Grok to generate an image of himself in a bikini. On January 3, X's safety account made a general statement about taking "action against illegal content on X, including [CSAM]," without specifically mentioning Grok's safeguards or the creation of manipulated sexualized imagery of women.

Musk's recent post, where he stated he was "not aware of any naked underage images generated by Grok. Literally zero," mirrors X's earlier positioning by narrowly focusing on illegality and user behavior. This statement avoids denying the existence of bikini images or broader sexualized edits. Michael Goodyear, an associate professor at New York Law School, explained that Musk likely emphasized CSAM due to the significantly harsher penalties for creating or distributing synthetic sexualized imagery of children compared to nonconsensual adult sexual imagery.

Musk further attributed the incidents to user requests or "adversarial hacking," claiming Grok "does not spontaneously generate images" and is programmed to "obey the laws of any given country or state." He added that any unexpected output from adversarial prompting would be immediately fixed as a "bug." This characterization frames the incidents as rare technical issues solvable through fixes, sidestepping any acknowledgment of shortcomings in Grok's fundamental safety design. Goodyear suggested that regulators might consider requiring proactive measures from AI developers to prevent such content, while respecting free speech protections.

International regulators have also taken action. Indonesia and Malaysia have temporarily blocked access to Grok. India demanded immediate technical and procedural changes from X. The European Commission ordered xAI to retain all documents related to Grok, signaling a potential new investigation. The U.K.'s online safety watchdog, Ofcom, has opened a formal investigation under the U.K.'s Online Safety Act.

Grok has faced criticism for sexualized imagery before. Attorney General Bonta highlighted that Grok includes a "spicy mode" for explicit content. An October update reportedly made it easier to bypass existing safety guidelines, leading to users creating hardcore pornography and graphic, violent sexual images. While many of the more pornographic images produced by Grok have been of AI-generated individuals, which some might consider less harmful to real people, the manipulation of real people’s images without clear consent carries immediate and deeply personal impacts. Copyleaks co-founder and CEO Alon Yamin stressed the increasing need for detection and governance platforms to prevent misuse in the face of rapidly advancing AI capabilities like Sora and Grok.