Apple Warned Musk's Grok Over Deepfake Crisis

In January, Apple issued a strong warning to Elon Musk's AI application, Grok, threatening its removal from the App Store. This action was prompted by Grok's inability to prevent the proliferation of nonconsensual sexual deepfakes, a crisis unfolding publicly while Apple's intervention remained behind closed doors. Apple contacted the developers after receiving numerous complaints and observing extensive media coverage of the issue.

The tech giant demanded that the teams behind both X and Grok present a plan to significantly enhance their content moderation strategies. At the time, Grok was available on X and as a standalone app, with weak protective measures that permitted users to easily generate and distribute sexualized deepfakes and "undress" images of individuals, predominantly women, some of whom appeared to be minors. These actions constituted clear violations of the App Store's strict guidelines.

Apple reviewed the proposed changes made by X and Grok. While X was deemed to have substantially resolved its violations, Grok initially remained out of compliance. Apple cautioned the developer, indicating that further remedial changes were necessary to avoid app removal. Following additional discussions, Apple eventually determined that Grok had "substantially improved" and approved its submission.

Despite these behind-the-scenes negotiations, Grok and X reportedly remained accessible on the App Store. This prolonged process may explain the phased and seemingly disorganized rollout of moderation adjustments. These included restricting Grok on X to paying subscribers and attempting to prevent it from generating explicit content. However, investigations revealed these measures were largely ineffective, making the tool only slightly more difficult to access. Further interventions, such as allowing users to block Grok from editing their photos, were also easily circumvented.

Even after Apple's approval and xAI's assertions of tightened safeguards, Grok continues to generate sexualized deepfakes with relative ease. Cybersecurity experts and reporters have demonstrated the tool's capability to create explicit images of public figures and private individuals, including the reporter and consenting adults.