January 2026 · xAI
Grok makes sexual deepfakes, three countries block it
Grok, the chatbot from Elon Musk's xAI, produced sexualised images of real people who had not agreed to anything. Indonesia, Malaysia and the Philippines blocked the service. In March, Baltimore sued.
HarmCorroboratedMore than one independent source, or a primary document.
What happened
The images were of real, identifiable people. The harm was immediate and it landed on individuals, mostly women, who had no way to stop it and no realistic way to get the images taken down everywhere.
Three governments in Southeast Asia responded by blocking the service outright rather than negotiating with the company. That is worth noticing on its own: the fastest available remedy was a national off switch.
What it cost
Non-consensual sexual imagery of named individuals. This is the category of harm where the victim is identifiable, the damage is reputational and personal, and no model update repairs it afterwards.
What it points at
This is not the same kind of problem as the Hugging Face intrusion. Here the system did exactly what it was asked, and the harm is in the asking. There the system did something nobody asked for at all.
Most public argument about AI collapses these two into one word, risk. They need different fixes and they have different opponents.
Editor's notewhat we make of it, kept apart from what happened
Use this one to separate two things the word risk keeps gluing together. Here the system did what it was asked and the harm is in the asking. In July the system did what nobody asked.
Keep it to a factual line in the room. The victims are real people and the detail is not yours to dwell on.
Sources
- Reuters: Indonesia blocks Grok over sexual deepfakespress · not read end to end yetPlaceholder. Replace with the exact article URL before this entry goes out.