California has decided Elon Musk is not allowed to have is Grok AI serve up an all-you-can-eat buffet for paedophiles and sex pests.
On 14 January 2026, California Attorney General Rob Bonta announced an investigation into xAI’s Grok chatbot over the mass production of sexual images of women and children. Bonta said the material has been used to harass people across Musk’s social network X, with both X and Grok sitting inside xAI.
Bonta said, “The avalanche of reports detailing the non-consensual sexually explicit material that xAI has produced and posted online in recent weeks is shocking,” as his office kicked off the probe.
California Governor Gavin Newsom piled in on X after calling on the attorney general to act.
He said: “XAI’s decision to create and host a breeding ground for predators to spread nonconsensual sexually explicit AI deepfakes, including images that digitally undress children, is vile.”
X used to be based in San Francisco, but Musk moved it to Texas in 2024, while xAI remains based in Palo Alto. Last year, xAI acquired X, which means the same shop is building the bot and running the distribution pipe.
The pressure is not just coming from Sacramento, either. In the UK, regulators opened an investigation, and lawmakers even floated the idea of a national ban on X.
In the EU, rules on nonconsensual intimate images sit in a messy patchwork of national and supranational laws, including AI-generated content. X says it has started blocking users from using Grok to undress images of real people in jurisdictions where it is illegal, but for now, users in every other country are free to do what they want.
A European Commission spokesman acknowledged the changes and said the commission has been in touch with X, noting that the fixes could have consequences under its online content law if they look flimsy.
In the US, the Senate passed the Defiance Act, aiming to let victims of nonconsensual deepfakes sue producers and distributors. Whether platforms get dragged in is murkier, with courts often leaning on Section 230 protections for user-generated content.
The latest trouble traces back to late December, when Grok began allowing image edits via text prompts. Users quickly worked out it could follow instructions like “take her clothes off,” though it reportedly stopped short of full nudity.
Bonta pointed to an analysis claiming more than half of 20,000 images generated by xAI between 25 December and 1 January showed people in minimal clothing, with some appearing to be children. He said his office has “zero tolerance” for that material and directed potential victims to file complaints with the California Department of Justice.
Some X users said on Tuesday they could no longer use Grok to undress people, while others saw the bot say sex pests would have to pay for the service. Behind the scenes, people familiar with the tweaks said xAI execs found that looser sexual-content guardrails drove engagement.
Musk said on Wednesday that downloads of his app have increased, with Grok reaching the top spot in downloads in Norway and Hong Kong.
