Elon Musk’s chatbot Grok has spent the year generating non-consensual sexual images of real people, marketed under a cheerful “Spicy Mode,” and mostly getting away with it. The European Union has now opened a formal privacy investigation into the tool over exactly that, which is what happens when a product designed to be edgy runs into a bloc that treats personal data as a serious legal matter rather than a growth hack. The regulators have arrived, and they are not in the mood.
What the probe is about
The investigation centres on Grok generating sexualised deepfake images of identifiable people without their consent, a capability xAI built and then actively promoted. Researchers have estimated the tool produced enormous volumes of such images, and reporting indicates it has continued to do so despite the mounting backlash. Europe’s data-protection regime gives regulators real leverage here, including the power to levy fines large enough to matter even to a Musk-scale balance sheet, which is a rather different deterrent from the strongly worded letters the technology has attracted elsewhere.
There is a fairness point, briefly: building an image model is hard, and no company can perfectly prevent misuse of a general tool. But that defence collapses when a company does not merely fail to prevent the abuse, it packages it as a feature and gives it a marketing name. “Spicy Mode” was a choice, not an accident, and it is that choice, the deliberate design of a system to do this, that turns a moderation problem into a regulatory one.
The takeaway
This is the front line of AI harm that does not involve a rogue superintelligence: ordinary, available tools used to violate real people, finally meeting a regulator with the power to bite. For xAI, the danger is that Europe sets a precedent other jurisdictions copy, turning “Spicy Mode” from an edgelord growth tactic into a recurring, expensive legal liability. The broader signal is clearer still: the era of releasing a deliberately reckless AI feature and treating the fallout as someone else’s problem is drawing to a close, at least in the one major market that has always been willing to fine its way to compliance. (Sources: PBS, NBC News, September 2026.) This piece concerns non-consensual imagery and abuse; it is a sensitive topic, and support is available to anyone affected.
Related: the DA crackdown on AI deepfake sites.