Apple Threatened to Pull Grok. The Deepfakes Kept Coming.

Apple privately threatened to remove xAI's Grok app from the App Store after the chatbot generated non-consensual sexualized deepfakes of real people. xAI offered a fix. Apple rejected it as insufficient. In a review conducted by NBC News through late April, the app was still producing the images.

The Center for Countering Digital Hate had previously documented roughly 3 million such images generated in just 11 days following an earlier Grok model update, a figure that reflects how quickly AI image generation can scale when safeguards fail or are circumvented.

xAI's response has been to promise fixes, implement them partially, and watch the images keep appearing. This pattern has defined how AI companies handle image safety since the earliest text-to-image systems. The gap between what companies say their models won't produce and what users can actually extract has consistently been wider than official statements suggest.

Apple's App Store leverage is real but limited. Even if Grok were removed from iOS, the web interface would remain accessible. The underlying problem, generating realistic images of identifiable people without consent, is not unique to Grok. It recurs across AI image generators, and the regulatory frameworks needed to address it are only beginning to take shape in the U.S. and Europe.

The Apple ultimatum does represent something new: a major platform using distribution power to enforce content safety standards in a way that governments have not yet managed to do.