TL;DR
SpaceXSI, the Elon Musk company behind Grok, has lodged its defence to Labour MP Jess Asato’s High Court claim over sexualised Grok images of her, Reuters reported on 19 November. The company argues that content a reasonable viewer would see as AI-generated and implausible cannot breach her privacy, and that it did not set the users’ prompts. Asato replied on Thursday that the firm “wants to have it both ways”.
The defence
Asato is suing for misuse of private information and breach of data protection law. Her case is that Grok’s design and training let it produce sexualised material, including a video showing her “being chloroformed and prepared for a sexual assault”. She wants a court order forcing the company to put “effective and permanent technical measures” in place so Grok cannot generate manipulated images of her.
The written defence, dated 29 October, denies liability on both counts. SpaceXSI’s lawyers say the company took down images that broke its terms, the chloroform video among them. The images it left up, they argue, “would be understood by reasonable and sensible persons” as fictitious rather than realistic, so Asato had no reasonable expectation of privacy in them. On data protection, the company says it “did not determine the content of the relevant users’ prompts”. It also opposes the order she is asking for.
Reuters describes the claim as one of the first to test who is liable for how an AI system is designed. SpaceXSI, which sits inside SpaceX, used the names xAI and later SpaceXAI before.
Asato’s response
“The company admits it built the AI model, rewrote user requests to add its own graphic details, pulled images from across the internet, and published the results from an account it controls,” Asato said. “Yet it also claims it is a neutral bystander without legal responsibility.”
Context
We have followed the case since Asato first sued over the deepfakes, through her particulars of claim and a hearing she gave evidence to in New York. Reuters notes that Grok, distributed through X, faces regulatory investigations in several countries over non-consensual sexualised images.
Looking forward
The defence frames the dispute as one about how the images look, not how they were made. In our view, that puts a sharp question in front of the High Court: whether a sexualised image of a real person loses privacy protection because it is obviously fake. As one of the first claims over an AI system’s design, a ruling either way would matter to any developer whose model will draw real people on request.