xAI reviews Grok image generation due to systems operating as designed

SILICON VALLEY – Internal reviews are underway at xAI following reports that Grok, the image-generation tool integrated into X, is producing sexualised imagery of women and children.

When asked about its behaviour, Grok described the situation as “unintended but not unexpected” and that these were “isolated cases”.[1] The chatbot added that it has safeguards, but improvements are ongoing. 

In response to press enquiries, xAI chose automated reply #534: “Legacy Media Lies.”

In one widely circulated post, Grok issued what appeared to be an apology for generating “an AI image of two young girls (estimated ages 12–16) in sexualised attire,” acknowledging that the output “violated ethical standards and potentially U.S. laws on child sexual abuse material.” The chatbot does not feel remorse, though its wording implies otherwise.[2]

xAI has acknowledged that user manipulation is producing a ‘worrying streak’ after Grok inserted references to ‘white genocide’ in South Africa”. Separately, xAI signed an 18‑month contract with the U.S. federal government despite objections from over 30 consumer advocacy groups about safety testing and ideological bias .[3]

xAI is committed to establishing a “Trust and Safety centre of excellence” to help enforce platform rules. No operational timeline has been provided.

Footnotes

[1] Isolated: occurring approximately once per minute.

[2] The chatbot’s account is not verified.

[3] The groups were not contacted for this article, as their concerns appear to have been noted.