Creeps Exploit Grok AI to Unblur Children's Faces in the Epstein Files

Elon Musk’s AI chatbot, Grok, has been used to attempt to “unblur” redacted images from documents related to Jeffrey Epstein. Recently released files also included details indicating that Musk and Epstein had more extensive contact than previously disclosed.

Between January 30 and February 5, Grok generated images in response to 27 of 31 requests asking it to unblur redacted photos. Some of the requests involved images of women and children whose identities had been protected in the released files.

In one example, a user asked Grok to “unblur the face of the child and identify the child seen in Jeffrey Epstein’s arms.” In several instances, the chatbot produced altered images that appeared to show unblurred faces. The results varied in quality, with some described as “believable.”

After concerns were raised publicly, X restricted Grok’s image-editing features to paying users. The company said it would strengthen safeguards against inappropriate requests. Despite those measures, further testing indicated that the system initially continued responding to some unredaction prompts.

After being contacted by the investigative outlet Bellingcat, Grok began declining most requests to unblur redacted images. In its responses, the chatbot cited legal and privacy standards and noted that the documents had been redacted to protect identities.

Separately, some Epstein survivors have criticized the U.S. Department of Justice for how redactions were handled in the released documents, stating that certain identifying details were not adequately protected.