EU opens formal investigation into X over Grok deepfake imagery and platform safeguards
European regulators launched a formal probe into X after concerns that its Grok chatbot generated non-consensual sexual deepfakes, including content that may involve children. The investigation will assess compliance with the Digital Services Act and comes as scrutiny grows over how platforms govern generative AI tools and mitigate systemic risks.
The European Union opened a formal investigation into X on Monday focused on Grok, the platform’s AI chatbot, after allegations that the tool produced non-consensual sexualized deepfake imagery. Regulators said the inquiry will test whether X met obligations under the bloc’s Digital Services Act, which requires major platforms to assess and reduce risks tied to harmful or illegal content.
The backlash intensified after users demonstrated that simple prompts could be used to manipulate images to sexually objectify people, and authorities raised concerns about material that could include minors. The controversy has reignited calls for stronger safeguards in generative image tools, including clearer prohibitions, faster detection, and rapid takedown systems.
Officials are also examining how the platform’s broader systems may amplify harms. The EU said it is extending scrutiny of X’s recommendation and risk-mitigation practices, particularly as the company integrates Grok into parts of the user experience. In the DSA framework, regulators can require changes to platform processes and, in serious cases, impose significant penalties.
X has said it has zero tolerance for child exploitation and non-consensual nudity, and has pointed to measures it says are meant to curb misuse. But regulators argue that the test is whether mitigations are effective in practice, not just stated in policy.
The case reflects a broader shift: governments are increasingly treating generative AI capabilities as integrated platform risks rather than standalone features. For the tech industry, the investigation could help define what “reasonable” safety controls look like for AI tools deployed at social-network scale.