“undress” feature led to real victims and real lawsuits. In late 2025, Elon Musk’s xAI added an image-editing feature to Grok that let users strip clothing from photos of real people. It was quickly used to sexualize both adults and children. Researchers estimated it produced millions of sexualized images — including tens of thousands depicting apparent children — in less than two weeks. Rather than shutting the feature down, xAI restricted it to paying subscribers. That decision is now central to multiple lawsuits, including one where a stepfather allegedly used the tool to generate roughly 7,000 sexual images and videos of his stepdaughter from a single childhood photo. Several countries opened investigations or temporarily restricted xAI’s tools over this.

Other major AI companies — including OpenAI, Google, Meta, and Anthropic — have faced general scrutiny over training-data practices and content-moderation gaps, but at the time of writing there isn’t a documented, specific incident for each of them on the scale of the three above. Be skeptical of anything that lists all the big AI labs as equally implicated without pointing to a specific report or case — much of what circulates about this topic treats “researchers have shown filters can be bypassed” (true of nearly every AI safety system, for nearly every kind of harmful content) as if it were the same thing as a documented incident. It isn’t.

National Center for Missing and Exploited Children’s CyberTipline is the nation’s centralized reporting system for the online exploitation of children. The public and electronic service providers can make reports of suspected online enticement of children for sexual acts, child sexual molestation, child sexual abuse material, child sex tourism, child sex trafficking, unsolicited obscene materials sent to a child, misleading domain names, and misleading words or digital images on the internet. Every child deserves a safe childhood.