FBI Subpoenas X for Grok Prompts Used to Create Deepfakes

Court records obtained by 404 Media reveal that the FBI subpoenaed X (formerly Twitter) for details about prompts used with its Grok AI chatbot to create more than 200 sexual deepfakes of a woman the suspect allegedly knew in real life.

The case highlights how law enforcement is increasingly targeting AI platforms and their data when investigating harassment and non-consensual intimate imagery. According to the court documents, the FBI sought information about specific prompts entered into Grok that were used to generate the deepfake videos.

This investigation represents one of the first known instances where federal authorities have formally requested data from a major social media platform about AI-generated content used for harassment. The subpoena demonstrates that law enforcement views the prompts and interactions with AI systems as potential evidence in criminal investigations.

The case raises questions about the responsibilities of AI companies and platforms in preventing misuse of their technology. While Grok is designed for general conversation and content generation, this incident shows how such tools can be weaponized for targeted harassment when used with malicious intent.

X has not publicly commented on the subpoena or the specific case. The company's policies prohibit the sharing of non-consensual intimate imagery, but the effectiveness of such policies when AI-generated content is involved remains an open question.

This investigation comes amid growing concerns about the proliferation of deepfake technology and its potential for abuse. Law enforcement agencies are still developing frameworks for addressing AI-generated harassment, and cases like this may help establish precedents for how such investigations proceed in the future.

The suspect in this case has not been publicly identified in the court records reviewed by 404 Media. The case underscores the challenges that both victims and law enforcement face when dealing with AI-generated harassment, where the technology can be used to create convincing but entirely fabricated content at scale.