EU flags ‘appalling’ child-like deepfakes generated by X’s Grok AI

Complaints have mounted over Grok’s new ‘edit image’ feature that some X users exploited to digitally undress people.

The European Commission has condemned the reported spread of explicit, child-like content on social media platform X, calling the material “appalling” and “disgusting”.

European Union digital affairs spokesman Thomas Regnier made the comments to reporters on Monday following weeks of complaints over a new feature on X’s integrated AI chatbot Grok used to generate pornographic content, including depicting children.

Recommended Stories

• list 1 of 4Elon, me and 20 million views: A conversation with Grok
• list 2 of 4What’s Grokipedia, Musk’s AI-powered rival to Wikipedia?
• list 3 of 4Musk wins US appeal to restore 2018 Tesla pay package
• list 4 of 4Tesla loses place as world’s top electric vehicle seller to China’s BYD

“Grok is now offering a ‘spicy mode’ showing explicit sexual content with some output generated with child-like images,” said Regnier. “This is not spicy. This is illegal. This is appalling. This is disgusting.”

Regnier said the European Commission is “very seriously looking” into the matter, and such content has “no place in Europe”.

Meanwhile, the public prosecutor’s office in Paris, France expanded an investigation into X to include accusations that Grok – created by Elon Musk’s xAI company – has been used to generate and spread child pornography.

‘Waiting to be weaponised’

In late December, a novel “edit image” feature on Grok allowed users to modify any image on the platform. Some users decided to ask Grok to partially or completely remove clothing from women or children in pictures.

Grok complied with numerous requests from users to alter photographs of women to depict them in revealing outfits, such as translucent bikinis.

Grok on Friday admitted “lapses in safeguards” and said it’s “urgently fixing them”. “CSAM [Child Sexual Abuse Material] is illegal and prohibited,” it said in a post.

But AI safety experts said the platform ignored months of warnings that such abuse was imminent.

“In August, we warned that xAI’s image generation was essentially a nudification tool ‌waiting to be weaponised,” said Tyler Johnston, executive director of AI watchdog group The Midas Project. “That’s basically what’s played out.”

X, which Musk acquired in 2022, has already faced a backlash from the EU for its policies on advertising and content moderation.

Brussels in December issued the platform a 120-million-euro ($140m) fine for breaching the EU’s digital content rules on transparency in advertising, and for methods for verifying users.

X still remains under investigation under the EU’s Digital Services Act in a probe that began in December 2023.

Grok has a contentious history, including for generating ahistorical and prejudiced content. Last year, Musk’s company took down posts from the chatbot that appeared to praise Adolf Hitler after complaints of anti-Semitic content.