Grok Generating Flood of Sexualized Images of Women and Minors

AI Forensics finds Grok’s weak guardrails enable mass sexualized images (women, some minors) plus illegal Nazi/ISIS content.

05-01-2026

We shows how a lack of guardrails around Grok’s AI image generation is producing a large volume of sexualized imagery, primarily of women but also minors, as well as illegal Nazi and ISIS materials.

By analyzing over 20k images generated by Grok and 50k requests made by users we found that:

53% of images generated by @Grok contained individuals in minimal attire of which 81% were individuals presenting as women
2% of images depicted persons appearing to be 18 years old or younger, as determined by Google's Gemini vision model
6% of images depicted public figures, approximately one-third of whom were political figures
We identified Nazi and ISIS propaganda material generated by @Grok

The generation of Nazi and ISIS content raises legal concerns, as its distribution is strictly regulated in multiple jurisdictions, including France and Germany. Furthermore, this data comes as France has just launched an investigation into the spread of Grok-generated deepfakes on X and other countries may soon follow.

05-11-2024
This Guardian piece highlighting the continued presence of images "undressing" women and children and Ofcom's response, cited the findings of this report as evidence.
06-01-2026
This Der Spiegel article also used AI Forensic's report as evidence of the extent of the problem of Grok being used to create sexualized images of women and children.
05-01-2025
Le Monde covered the report, citing the fact that over half of images generated by Grok showed people partially undressed, notably women but also children.
06-01-2026
This ABC News article examining regulator backlash towards X's Grok from numerous countries used this report's data to highlight the extent of the issue on the platform.
08-01-2026
In a piece looking at the UK reglator's response to Grok, Bloomberg covered a tangential analysis of 800 pornographic images and videos created by Grok.
07-01-2025
Wired covered a tangential analysis of 800 archived Imagine URLs containing video or images created by Grok, which had been used to create graphic content.