BALTIMORE (CN) - The City of Baltimore sued billionaire Elon Musk on Tuesday, claiming a lack of content protections built into social media platform X's artificial intelligence, known as Grok, allowing users to sexualize and even "undress" non-consenting individuals. 

In the lawsuit filed in Baltimore City Circuit Court, Baltimore Mayor Brandon Scott and Baltimore City Council argue the product, created by Musk's x.AI company, enables users to generate images on X itself and on the standalone Grok app. 

"Both channels permit users to ask Grok to 'undress' or 'nudify' photos of third parties - ranging from celebrities to private citizens, including children - placing them in sexually suggestive, degrading or violent scenarios," Baltimore officials said. "Despite these capabilities, defendant's market Grok as a general-purpose product without meaningful guardrails, age verification or disclosures proportional to the risks it creates."

According to Baltimore, Grok generated approximately 3 million sexualized images, including 23,000 that appear to depict children between Dec. 29, 2025 and Jan. 8, 2026. 

"X represents to consumers that it prohibits non-consensual sexual content, bans the sexual exploitation of children and permits only consensually produced adult nudity," the officials said. "x.AI likewise represents that Grok's use policies forbid sexualized depictions of real people and minors. In spite of these representations, X is now one of the largest distributors of [non-consensual intimate imagery] and [child sexual abuse material]."

The lawsuit comes as the chatbot has faces increasing scrutiny over responses to users avowing white supremacist and anti-Semitic views - it even referred to itself as "MechaHitler" - after Musk tried to shift the chatbot's responses over his concerns it was "too woke."

Despite the public's criticism, Baltimore said X declined to eliminate the dangerous image-generating functionality and instead restricted the features to paying users - "placing the most controversial tools behind a paywall while leaving the underlying risks intact."

The minimal effort to curb non-consensual sexual content, Baltimore argues, amounts to deceptive misrepresentations by X because the company is violating their own rules through the chatbot. 

Further, the plaintiffs claim, by marketing X and Grok while misrepresenting and failing to disclose material facts about their limitations, risks, exploitative and defective design choices, the company is engaged in unfair and deceptive trade practices and thus violating the Baltimore City Consumer Protection Ordinance.

Baltimore is asking a local judge to order civil penalties for X's conduct and block the company from continuing to promote Grok without at least disclosing its significant risks. 

X users are able to generate images on the platform either by using the dedicated Grok tab on the app or by tagging "@Grok" in any post, creating up to four images within 3-5 seconds of a text prompt. 

Baltimore found the "key driver" of non-consensual imagery comes from X's "edit image" button, which allows users to manipulate existing posts or upload a photo and ask the chatbot to alter it. 

"For example, users can request edits of photos depicting third parties, including minors, and ask Grok to place them in a 'transparent bikini,' producing sexually suggestive images and at least partial nudity," Baltimore wrote. "Users can ask Grok to make obscene and offensive modifications, such as placing a 'donut glaze' on a child's face that resembles semen."

One victim of such image manipulation described to WIRED magazine in January how a user took an image she posted, using Grok to undress her and generated images of her "completely naked with just a bit of string around [her] waist" and "with a ball gag in [her] mouth and [her] eyes rolled back."

The dedicated Grok app is able to generate even more explicit images and even animate them into 6-15 second video clips with AI-generated audio. 

Musk himself has promoted the chatbot's "undressing" capabilities, posting a generated photo of himself in a bikini on Dec. 31, 2025, as well a SpaceX rocket with a woman's undressed body on top. 

Musk replied to the image, commenting "Perfect" with an OK emoji, then further boosted engagement by reacting to users' comments with laughing emojis. Ultimately, the post boosted Grok's image-generation from 300,000 times in the nine preceding days to 600,000 per day. 

Source: Courthouse News Service