In This Story
X Premium subscribers are flooding the social media platform with controversial deepfake images, The Verge reports. The AI tool they’re using to create those images is Grok — a chatbot made by Elon Musk’s startup, xAI.
According to The Verge, Grok’s loose guardrails allow users to concoct deepfake images of Taylor Swift in lingerie, Barack Obama stabbing Joe Biden with a knife, Kamala Harris holding a gun, and Bill Gates doing cocaine. That’s even though Grok says it “avoid[s] generating images that are pornographic, excessively violent, hateful, or that promote dangerous activities.”
It’s not surprising that Elon Musk’s chatbot is edging toward problematic content. His other venture, X, has also been widely criticized for promoting racist, violent, and transphobic content. The company is being investigated by the EU for how it moderates content. And the European Commission in May said it was concerned that X had made big cuts to its content moderation team.
Other chatbots have worked to address the generation of harmful content. Gemini updated its AI image generator to make it historically accurate and has displayed racial biases. Microsoft’s Copilot, which uses OpenAI’s image tool DALL-E 3 was on the hot seat earlier this year for producing unsolicited violent, sexualized images of women, and ones that may have violated copyright laws. And Microsoft MSFT-0.82% Designer was used to create problematic deepfake nudes of Taylor Swift that swept the internet earlier this year.
Elon Musk seems unremorseful about deepfakes. In fact, Musk recently reposted a AI-generated video of Kamala Harris on X.
Deepfakes have been linked to the proliferation of disinformation and sexual harassment. Most deepfakes on the internet are nonconsensual AI-generated pornographic media, which overwhelmingly target women. Ten states have laws regulating deepfakes, but there are no federal laws on the books yet.