Pull to refresh

Cover up

Under growing pressure, Elon Musk backs down on nude deepfakes

January 15, 2026

 Elon Musk
Editor’s note (January 15th 2025): This article has been updated to take in developments.
FROM DRUG use to libel suits, Elon Musk has long been able to shrug off problems that would have imperilled most bosses. But the reaction to a recent update to Grok, the AI chatbot linked to X, the billionaire’s social network, has posed a fresh test of his imperviousness. Grok’s willingness to churn out nude deepfakes has led to scrutiny in Britain and Australia, among other places, and outright bans of the chatbot in Malaysia and Indonesia.
The flood of deepfakes began after December 24th, when Mr Musk said that Grok had been updated with a new image generator. The change brought the chatbot to parity with those of Google and OpenAI, but with some big differences. In line with Mr Musk’s desire to create the most anti-censorship, “maximally truth-seeking” chatbot, the system is far less likely than others to refuse requests to generate images that violate content policies. Ask ChatGPT to edit a picture of a real person to strip them to their underwear, and it will demur. Grok has no such qualms.
Moreover, because Grok can be accessed on X, it answers many requests in public. Until another update limited the feature to paying subscribers on January 9th, the chatbot’s public profile was responding to a hundred such “nudification” requests a minute, some of them involving children. The phrase “put her in a bikini” soared on the social-media site (and some users have managed to go further still, generating pornographic images with cleverly worded prompts). Other companies might have declared such creations an abuse of their product and apologised; Mr Musk has responded with laughter to an image putting a toaster in a bikini.
The fallout is proving harder to laugh off. Indonesia, home to 24m X users, has responded with a temporary ban of Grok, calling the imagery “a serious violation of human rights, dignity and the security of citizens”. Malaysia, with another 5m users, has followed suit. In Britain, the Labour government has announced plans to accelerate the enforcement of new laws against nude deepfakes, passed by parliament last summer, alongside an investigation by Ofcom into various breaches by X of the Online Safety Act, including failures to take down illegal content such as child-sexual-abuse material swiftly. A fine from Ofcom would be noticeable, but not a disaster for the deep-pocketed Mr Musk. For serious breaches of the Online Safety Act, the regulator can levy penalties of up to 10% of a firm’s global revenue—around $250m in X’s case.
The bigger question is what happens in America, X’s largest market. Last May Donald Trump signed into law the “Take It Down Act”, which penalises platforms that spread “intimate visual depictions” of a person online, giving companies a year to prepare. But some of the images causing controversy today might not meet the definition of an “intimate visual depiction”, says Riana Pfefferkorn of Stanford University’s Institute for Human-Centred AI. Nor is enforcement guaranteed. X benefits from its owner’s close links to the Trump administration. After the EU issued fines against the social network in December for violations of its Digital Services Act, America’s secretary of state, Marco Rubio, warned that “the days of censoring Americans online are over.”
Even so, pressure in America is mounting. On January 13th the Senate passed a bipartisan bill that would allow victims to sue over non-consensual deepfakes. On January 14th California’s top prosecutor said it was investigating the sexualised deepfakes produced by Grok. Then there are the commercial risks. Many advertisers are already wary of X. Payment processors such as Visa and Mastercard have strict content standards and have shown a willingness to enforce them—such as in 2021, when they required pornographic websites to verify the age of anyone featured in online content. All that helps explain why, on January 14th, Mr Musk appeared to back down, with X announcing that Grok would no longer be able to produce nude deepfakes of real people in “jurisdictions where it is illegal”. Even the world’s richest man must occasionally bow to pressure.
To track the trends shaping commerce, industry and technology, sign up to “The Bottom Line”, our weekly subscriber-only newsletter on global business.