Grok’s Sexualized Deepfakes Violate Canada’s Privacy Law, Watchdog Rules

OTTAWA — Users of Elon Musk's Grok AI chatbot generated more than 1.8 million sexualized deepfake images worldwide in just 10 days earlier this year, according to Canada’s privacy commissioner, who says X and xAI continue to violate federal privacy law by allowing the creation and sharing of nonconsensual sexualized images.

Privacy Commissioner of Canada Philippe Dufresne released his findings Thursday, saying that despite bringing in new safeguards, X and xAI continue to allow users to create nonconsensual sexual deepfakes using Grok.

The ruling adds pressure on social media and artificial intelligence companies to curb the problem as Canada moves to criminalize the sharing of the images and impose new requirements on online platforms to remove them.

“They have made some commitments to bring themselves in compliance, but in this case, I was not satisfied that this issue is resolved,” Dufresne said.

If X does not comply, Dufresne said he can ask the federal court of Canada to enforce the law, but it’s a lengthy and expensive process.

The two companies rejected his recommendation to pause Grok until the issue is fixed, Dufresne said, but they have committed to providing his office with regular audit reports until he is satisfied there are no longer nonconsensual deepfakes being shared.

Users on X have been prompting Grok to superimpose bikinis, lingerie and other revealing clothes on adults and children, prompting responses from governments and prosecutors around the world. Malaysia and Indonesia even temporarily banned the app; California and the U.K. have also launched investigations.

Dufresne's findings highlight media reports that at one point Grok was generating more than 6,000 sexualized deepfakes every hour — almost 2 million sexualized images since Dec. 29, 2025.

While a second study from the nonprofit Center for Countering Digital Hate estimated that between Dec. 29 and Jan. 8, Grok generated approximately 3 million sexualized deepfakes, including more than 23,000 images of children.

“Grok has highlighted they’ve since reduced the numbers by 50 percent. So it's a big reduction, but that means there's still many that are there,” Dufresne said.

“That’s why I’m not at this moment satisfied that it is resolved.”

X did not respond to a request for comment.

The privacy commissioner launched his investigation in January following other investigations in the U.K. and California.

Canada is also close to criminalizing the sharing of nonconsensual sexual deepfakes. Legislation that is expected to become law this month would make it a crime to share or threaten to share them and would require social media platforms to report AI-generated child sexual abuse material to police.

The Liberal government also introduced the Safe Social Media Act on Wednesday, a bill that would require platforms to remove sexually explicit content, including AI-generated deepfakes, within 24 hours.

In January, X announced it would stop users, including paid subscribers, from editing images “of real people in revealing clothing such as bikinis,” insisting its social media platform has a “zero tolerance” policy “for any forms of child sexual exploitation, non-consensual nudity, and unwanted sexual content.”