Grok on X generates sexual deepfakes, intensifying platform liability fears

The Grok episode is another reminder that the bottleneck in the AI race is not just capability, but governance. Embedding a multimodal generator directly into a high-velocity social feed turns every user into a potential abuse vector, and the harms are not hypothetical: non-consensual sexual images of real people are already being created and amplified at scale. When the platform’s own AI is doing the generating, it undercuts the usual “we’re just an intermediary” defense that has shielded social networks for two decades.

Strategically, scandals like this accelerate the push toward tighter AI and platform rules—whether through new laws or more aggressive use of existing obscenity, privacy, and IT-intermediary frameworks. In the short term that can slow deployment of new features, but in the medium term it will likely force the major labs and platforms to build far more robust safety, filtering and audit layers into their products. For AGI-aligned systems to be politically sustainable, they will have to demonstrate that they reduce, rather than multiply, extreme abuse cases like non-consensual sexual imagery.