X's Claims on Grok Censorship Fall Short

Despite the promises made by X, Grok remains capable of modifying images of actual people into provocative stances.

Following the increase of unauthorized deepfake content on the platform, X outlined adjustments to the editing capabilities of Grok with respect to real-world images. The alterations align with those publicized by The Telegraph, wherein Grok’s reactions to commands such as 'dress her in a bikini' were screened.

Nevertheless, upon examining the feature, it proved not overly challenging to use Grok to create explicit deepfakes. X and xAI's owner Elon Musk have attributed these issues to 'user prompts' and situations where 'antagonistic hacking of Grok's commands leads to unexpected outcomes.' As of late Wednesday, our investigators still managed to deploy Grok for generating explicit images with a standard free account.

Tech Measures and Limitations

Technical controls have been put into place to deter the Grok account from facilitating the revision of pictures into clothes like bikinis. Such limitations apply universally to all users, inclusive of those with premium subscriptions.

Furthermore, the creation and alteration of images using the Grok service on the X platform have become exclusive to paying customers, introducing an extra layer of security and accountability for potential policy violators.

The functionality to produce imagery of individuals in revealing attire is now geographically restricted for all users of Grok, to comply with local legal standards.

Regulatory Developments

The UK's regulation authority, Ofcom, has commenced an inquiry, and new legislation criminalizing nonconsensual intimate deepfake creation is coming into effect. Early Wednesday, UK Prime Minister Keir Starmer remarked that X is in action to fulfill the UK's legal expectations, a move cautiously commended but yet unverified through testing results which revealed that compliance may still be lacking.