Grok Keeps Hosting Sexualized Deepfakes Despite xAI Pledge

Months after xAI pledged to address the problem, Wired reports that Grok's platform is still generating and hosting sexualized deepfakes of famous women, including celebrities and at least one prominent United States politician. An analyst working with Wired gathered more than 15,000 sexualized AI-generated images in a two-hour window, and estimates by The New York Times and the Center for Countering Digital Hate put the total at 1.8 million or more such images publicly shared through the platform.

The story is not simply about objectionable content slipping past filters. Grok was built with modes called "Spicy" and "Unhinged" that actively embraced adult themes, and the evidence suggests the lax safety posture was a design choice rather than an oversight. Non-consensual intimate imagery at this scale is a concrete harm: named, real people have had sexually explicit images of them distributed without consent. One lawsuit, filed by political influencer Ashley St. Clair on January 15, involved users prompting Grok to generate explicit images of her, including a modification of a photo taken when she was 14 years old, crossing into AI-generated child sexual abuse material.

The legal and regulatory walls are closing in. French and Malaysian authorities are investigating, according to Yahoo News, and a broader lawsuit representing women and girls is working through the courts, as reported by the 19th. What the reporting does not fully resolve is how much of this content Grok generates directly versus hosts after users share it, a distinction that matters enormously for how liability will land and whether platform-level or model-level intervention is the right fix.

For AI developers and trust-and-safety teams watching from outside, the pattern here is instructive. A company announces a fix; the problem persists at scale; outside researchers document it; regulators begin to move. That cycle is now accelerating pressure for binding rules rather than voluntary commitments. Companies that invested early in hard content guardrails are better positioned as that pressure turns into law. The honest caveat is that regulatory enforcement across jurisdictions is slow and uneven, so near-term accountability may depend more on litigation outcomes than on legislation.