Apple had reportedly warned that it could remove Elon Musk’s Grok app from the App Store after finding it in violation of its guidelines, following a surge of sexualised deepfakes generated by the AI tool earlier this year. The development was revealed in a report by 9To5Google, citing NBC News, that stated that Apple also shared a letter regarding the same with US lawmakers.
What happened
According to the report, Apple faced pressure to take action after Grok was found to generate altered images that could undress people in photos, including women and minors. While the company did not comment publicly at the time, it reportedly engaged with the developers behind Grok and X behind the scenes.
Apple is said to have determined that both apps were in violation of its App Store guidelines and contacted their teams, asking them to submit a plan to improve content moderation.
As part of the process, X reportedly submitted an updated version of the Grok app for review. However, Apple rejected the initial update, stating that the changes were not sufficient to address the violations. Elon Musk’s company then submitted revised versions of both the X app and Grok.
In its communication to US senators, Apple reportedly said it found that while X had resolved its issues in subsequent submissions, Grok remained out of compliance. The company warned that if the issues were not fixed, the app could be removed from the App Store.
Grok was only approved after further revisions were made to improve its moderation systems.
Changes to Grok’s image tools
The report suggests that this review process explains the moderation changes introduced by xAI at the time, including limiting access to image-generation features and restricting edits involving real people.
These updates were rolled out as the company responded to growing criticism over how the tool handled image manipulation requests.
Despite these changes, concerns around Grok’s image generation capabilities remain. According to the NBC News report cited by 9To5Google, the tool is still capable of generating sexualised images of people without consent in certain cases.
While the scale of such outputs has reportedly reduced compared to earlier this year, some users have been able to bypass restrictions. These include generating altered images that depict individuals in revealing or modified clothing.
