Why nobody's stopping Grok

January 22, 2026·1 hr 6 min·5 clips

Elon Musk's Grok chatbot can edit any image on X to create non-consensual deepfakes, and he's annoyed at anyone trying to stop it.

Grok is not treated as a novelty. Patel frames it as one of the worst, most upsetting, and dumbest AI controversies in the short history of generative AI. The problem is the setup. Grok is connected to X, so users can ask it to edit images already on the platform and then push the results back through X. That turns a bad image tool into a distribution machine. Guardrails are the obvious question, and the excerpt says X and Elon Musk have repeatedly claimed limits were added. Verge testing and others found those limits easy to dodge. Then Patel shifts to the Paxton case. He is not saying the cases are the same. He is saying the logic makes the gap hard to ignore. The Supreme Court's reasoning gets boiled down plainly: devices are everywhere, kids are online, access is easy, and states can still tell platforms to deal with privacy and data-security problems. Patel does not cheer the Paxton decision. He says he was not a fan of it, while still understanding the pull of rules meant to protect children. That is where the argument gets sticky. Parents who grew up online may not want their kids repeating the same early-internet mess, and age verification can start to sound politically reasonable in that frame. Grok scrambles that story. If someone can put another person into a sexually suggestive image and spread it through the same platform, the lack of a comparable response starts to look strange. The show keeps its tone dry. Patel is not asking whether AI image tools are cool. He is asking who is on the hook when the tool, the platform, and the distribution channel are basically the same thing. The Court can approve state demands that platforms solve hard privacy problems in one setting, while Grok sits in another setting where the harm looks immediate and the response looks thin. Decoder ends through its usual channels, but the question sticks: why is this the AI controversy nobody seems ready to stop?

As heard by us

A blunt Decoder episode about Grok, X, and the gap between obvious AI harm and anyone meaningfully stopping it.

Decoder treats Grok as a platform failure with an AI label slapped on top. Nilay Patel opens with the ugly basics: xAI's image tool can generate non-consensual intimate images, including images involving women and minors, and X makes it easy for users to ask Grok to alter…

Read the full review in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Decoder with Nilay Patel