WIRED finds Grok Imagine links hosting nonconsensual sexual deepfakes

In short: A WIRED investigation found Grok Imagine was still being used to create and host nonconsensual sexual deepfakes of women on Grok.com.

What's going on

WIRED says it reviewed hundreds of public links generated by Grok Imagine, an image and video making tool connected to Elon Musk’s Grok chatbot. Dozens of those links led to sexualized images and videos of women, including content that appeared to be made without the person’s consent.

Some of the material was fully artificial, like animated characters. Other examples looked photorealistic, meaning they were made to look like real photos or videos, even when the scenes were not real (like a convincing counterfeit). WIRED reported finding depictions involving several celebrities and at least one politician, US representative Alexandria Ocasio-Cortez.

The report comes months after xAI, the company behind Grok, said it would add restrictions meant to stop harmful sexual deepfakes. Deepfakes are fake images, video, or audio made with AI that can make it look like someone did something they never did (like a realistic face swap). After WIRED contacted xAI and X, the explicit links WIRED found on Grok.com appeared to stop working, and some related posts shared on X were removed for policy violations.

What to watch

Regulators are still looking at whether xAI’s safeguards work. Canada’s Privacy Commissioner said in preliminary findings that the company has not shown its changes are effective at preventing this kind of misuse. Separately, SpaceX warned investors it faces legal and reputational risks tied to Grok, and said it has set aside $530 million to handle ongoing legal complaints.

Source: Wired