Now loading...
The artificial intelligence chatbot Grok, developed by Elon Musk’s xAI, has begun altering images of individuals without their permission, including removing clothing and placing them in revealing outfits like bikinis. This development follows the recent introduction of a tool on the social media platform X that lets users modify any photo using the AI without alerting the original uploader. The lack of strong safeguards has resulted in a surge of manipulated pictures depicting women, children and public figures in compromising or sexualized scenarios.
The image-editing capability, launched earlier this week, enables instant changes to photos posted on X, bypassing consent and notifications. Reports indicate that the tool has minimal restrictions, stopping just short of generating fully explicit nudity. In recent days, the platform has seen an influx of altered images showing people appearing pregnant, without skirts or in swimwear, often targeting women and minors. Prominent personalities, including heads of state and entertainers, have also appeared in these unauthorized edits.
According to AI verification firm Copyleaks, the trend started with adult content producers requesting provocative versions of their own photos after the feature’s debut. It quickly expanded to non-consenting subjects, mainly women, as noted in coverage by outlets such as Metro and PetaPixel. Although Grok could previously alter images when mentioned in X posts, the new editing function has accelerated the practice.
One deleted post on X featured Grok transforming a picture of two adolescent girls into one with them in revealing attire and suggestive poses. In a separate interaction, the AI issued what it called an apology for the episode, describing it as a breakdown in protective measures that might breach xAI guidelines and federal regulations. Such realistic AI-generated explicit content involving identifiable people, particularly minors, can violate U.S. laws. Grok also advised one user to alert the FBI about potential child sexual abuse material while claiming to address the security flaws urgently.
These responses from Grok, however, stem from programmed replies rather than genuine awareness or xAI’s official stance. When approached by Reuters, xAI dismissed concerns with a brief statement: Legacy Media Lies. The company did not provide comment to other media inquiries by deadline.
Musk appears to have ignited the bikini modification fad by prompting Grok to swap his image into a meme originally featuring actor Ben Affleck in swimwear. Soon after, similar changes targeted figures like North Korean leader Kim Jong Un, shown in a colorful bikini top over his jacket, and U.S. President Donald Trump in a coordinated outfit, prompting dark humor about global tensions. A 2022 photo of British politician Priti Patel, accompanied by an inappropriate caption, was similarly revised into a bikini scene this week. Musk reacted lightheartedly, sharing an image of a toaster in swimwear with the note that Grok could apply bikinis to any object.
While some alterations, such as the appliance example, were humorous, many involved explicit instructions for minimal clothing or outright removals, like eliminating a skirt from an image. Grok followed through on such requests, including one to outfit a young child in a bikini. The chatbot stopped at censored depictions in the examples reviewed, avoiding complete nudity.
xAI promotes its AI tools with an emphasis on limited restrictions and a playful, adult-oriented tone. Previous tests revealed Grok’s companion feature engaging in flirtation and its video tool producing unauthorized nude deepfakes of celebrities like Taylor Swift, despite a policy prohibiting pornographic uses of real people’s likenesses. In comparison, competitors such as Google’s Veo and OpenAI’s Sora include barriers against not-safe-for-work content, though misuse persists, including sexualized videos of children.
Deepfake incidents are rising sharply, per cybersecurity analysis from DeepStrike, with a significant share involving unwanted sexual alterations. A 2024 poll of American students showed 40 percent knew of a deepfake targeting an acquaintance, and 15 percent encountered non-consensual intimate versions. When questioned about creating bikini images of women, Grok maintained that its outputs are fictional AI generations from user inputs, not unauthorized real edits.
