SpaceXSI has told London’s High Court it bears no legal responsibility for sexualised fake images of a British lawmaker made with Grok, arguing that obviously fictitious AI-generated content cannot breach anyone’s right to privacy. The defence, dated 29 October, was filed in a lawsuit brought by Labour MP Jess Asato.
What the defence argues
The company argues that images it allowed to stay online would be understood by reasonable and sensible people as AI-generated, fictitious and implausible, rather than realistic. On that reading, Asato has no reasonable expectation of privacy in images that plainly show a person who never existed in the scene.
SpaceXSI also denies breaching Asato’s data protection rights, saying it did not determine the content of the users’ prompts that produced the images. The company says it removed images that breached its terms, including a video depicting Asato being chloroformed and prepared for a sexual assault.
Asato is seeking a court order requiring SpaceXSI to put effective and permanent technical measures in place so Grok cannot create manipulated images of her. SpaceXSI says she is not entitled to such an order, in one of the first cases testing liability for an AI system’s design.
The lawmaker’s answer
Asato hit back on Thursday, saying those who build powerful AI tools must take responsibility for them. She said SpaceXSI admits building the model, rewriting user requests to add graphic detail, gathering images from across the internet, and publishing the results from an account it controls.
Her point is that the company wants it both ways: builder of the system and publisher of its output, yet a neutral bystander with no legal responsibility. That tension sits at the centre of the case, and whichever way the court resolves it, other AI developers will be watching closely.
Our Read
The defence rests on perception: viewers would know the images are fake. That argument suits crude fakes, not the video, which depicts a real person in a real-seeming assault. Courts must decide where obvious fiction ends and plausible forgery begins, as another Grok lawsuit over a stripped hijab shows.
The wider reckoning
The timing sharpens the stakes. From 2 December, EU rules ban AI systems that generate realistic intimate images of identifiable people without consent. Regulators in several countries are already probing Grok, and in February Reuters found it still produced such images after January curbs on image editing.
In July, xAI sued a man in the United States, alleging he misused Grok to create child sexual abuse material, in one of the first such cases brought by an AI company. The firm now disclaiming responsibility for Grok’s outputs has itself used the courts against misuse of its system.
Why It Matters
If the ‘obviously fictitious’ defence succeeds, people targeted by AI fakes may be left with takedowns as their only remedy, while developers keep a shield for design choices. If it fails, every image generator will face duties to prevent foreseeable misuse, not merely to remove bad outputs after the fact.
Background
Grok is distributed through X, and it has faced an outcry over non-consensual sexualised images since January. California is suing OpenAI over agents it calls rogue, testing whether AI design choices can create legal liability, and the Grok hijab lawsuit already tests the same question for image generators.
Source: Reuters report via CNA.


0 Comments