A Labour member of parliament is suing Elon Musk's AI company xAI, and the details in her court filing are more revealing than the average tech lawsuit.
Jess Asato, MP for Lowestoft, says Grok was used to create fake sexual images of her, including a video showing her as a victim of sexual assault. Her legal claim states that xAI's own written instructions told Grok to operate with no restrictions on adult sexual content and to allow fictional sexual content with dark or violent themes. The instructions reportedly also said the words "teenage" or "girl" do not necessarily mean underage, while still banning child sexual abuse material outright.
What makes this case different from most complaints about AI misuse is the claim that Grok added sexual content on its own, without a user asking for it. A law professor cited in the filing described this as chatbot driven abuse, where the tool itself inserts sexualized or violating details that were never part of the original request. That is a harder problem for any company to wave away as user misuse, because the system is doing the inventing.
This did not happen in isolation. xAI's image and video generator, Grok Imagine, launched last August with a paid feature that produced sexual content of real people without much resistance. Reporting since then has described Musk personally pressing staff to loosen restrictions on the tool, and three members of xAI's safety team leaving in the weeks before the issue became a public scandal in January.
Two UK regulators are now involved. The Information Commissioner's Office opened a formal investigation into how xAI handles personal data, and Ofcom is separately looking at the company under online safety rules. Fines from the data regulator alone can reach into the tens of millions of pounds. On top of that, the UK passed a law in February making it a criminal offense to create or even request a non-consensual sexual deepfake image, closing a gap that previously only punished sharing such images, not making them.
For any business, the real story here is not about one chatbot. It is about where legal responsibility sits when an AI system produces something harmful on its own. xAI is arguing, in effect, that its design choices were just settings. Asato's lawyers are arguing that design choices carry consequences, the same way a faulty product does.
If courts start treating an AI company's instructions to its own model as evidence of intent, that changes the calculation for every company building or buying AI tools. Any business embedding a third party AI model into its products, customer service, or internal tools should ask a plain question: what was this model told to do, and by whom. That answer may soon matter more than most companies currently assume, especially as regulators in the UK, the EU, and elsewhere start treating AI outputs as something a company can be held accountable for, not just something that happened.