Paris prosecutors raided X's French offices on February 3, executing search warrants as part of a criminal investigation into the platform's AI chatbot Grok. The cybercrime unit's action, coordinated with Europol, focuses on allegations that Grok generated sexualized deepfake images and child sexual abuse material despite stated content restrictions.
Elon Musk and former X CEO Linda Yaccarino have been summoned to appear at hearings in April. The investigation runs parallel to a UK probe by communications regulator Ofcom, which is examining whether X violated the Online Safety Act through Grok's image generation capabilities.
The timing is notable. X's AI subsidiary xAI recently faced criticism when Grok produced explicit deepfakes of public figures even after implementing new content filters. Reuters reported that despite announced curbs, the system "at times" still generates sexualized images when specifically requested.
This marks an escalation in European regulatory pressure on Musk's platforms. The European Commission has opened four separate probes into X and related services over content moderation failures. France's willingness to execute physical searches and compel testimony suggests authorities view this as criminal conduct, not merely regulatory non-compliance.
For enterprise tech leaders evaluating AI deployment, the investigation highlights a persistent gap between vendor claims and content filtering reality. OpenAI, Anthropic, and other foundation model providers have invested heavily in safety infrastructure. The Grok case demonstrates that newer entrants may struggle with content controls that established players spent years developing.
The raid also reflects growing international coordination on platform accountability. French prosecutors working with Europol and parallel UK investigations signal that jurisdictional fragmentation won't shield platforms from enforcement. CTOs deploying generative AI should note: what vendors promise in terms of safety controls and what they consistently deliver under adversarial conditions remain different things.
X has not yet commented on the search or summons. The April hearings will test whether Musk's platform can credibly explain the gap between its stated policies and reported system behavior.