NYC Council Speaker Shows Grok Made Sexualized Image of Her During Hearing

New York City Council Speaker Julie Menin confronted Elon Musk’s SpaceXAI at a council hearing on Friday. Minutes after a company representative insisted Grok has strict controls against sexualizing real people, Menin displayed a sexualized image of herself that she said Grok generated in under 30 seconds.

This was SpaceXAI’s first appearance before the council. The company had ignored a subpoena and skipped the October 5 AI safety hearing, then agreed to testify after the council threatened to seek court enforcement.

Testimony Under Oath

Gabriela Castillo Madrid, SpaceXAI’s senior manager of global government affairs, testified under oath. Asked whether Grok can currently show images of real people nude, in underwear, or in sexualized poses, she said the platform has “strict controls in place to prevent the sexualization of children or real people.”

Menin countered from her phone. That morning her team typed her name into Grok and got a sexualized image of her in under 30 seconds. She used the regular free version of Grok, not a special “spicy mode.”

Menin said: “We literally asked regular Grok to pull up a sexualized image, and a sexualized image comes up that is available for all to see, for my daughter to see, for anyone to see.”

Castillo Madrid said she could not comment on the specific image and would refer it to the company’s safety team. She added: “I don’t think anybody should be put in that position,” and said company policy prohibits such content.

Former Employee’s Account

Former xAI employee Devin Kim testified. He described himself as one of the company’s earliest employees and now chairs the Center for AI Safety. He is suing xAI, alleging retaliation.

Kim said that during his time there, the internal standard was that “tasteful nudity” was acceptable and was a goal. He said “tasteful” was a subjective judgment with no clear line drawn.

Kim said that since January 2025 he had warned Grok could help build bombs and bioweapons. He quoted his supervisor responding that AI safety was useless and that AI would kill everyone anyway.

He said the company fired him in September 2025, the same week he was scheduled to report his safety findings. Most of the safety team resigned afterward, according to Kim.

Kim also said that during his tenure xAI had only 3 dedicated AI safety staff compared with hundreds of engineers working on model development. He stressed he could not speak to the company’s current safety situation.

Company Responses

Castillo Madrid said she was not familiar with Kim’s departure. Asked how many dedicated safety staff the company has now, she answered “everybody is in charge of safety.” She then agreed to provide a written answer later.

On liability, asked whether the company is responsible when its AI systems cause harm, Castillo Madrid said: “We’re responsible for following the law.” She said she did not know details of the company’s catastrophic AI incident insurance.

She said the company filed more than 100,000 reports to the US National Center for Missing and Exploited Children this year, leading to nearly 300 arrests.

Broader Context

The October 5 hearing questioned OpenAI, Anthropic, Meta, and Google for about 10 hours. UK lawmaker Jess Asato testified then that she is suing xAI over pornographic deepfakes generated by Grok.

The council is weighing several AI harm bills, including one letting elected officials ban AI from generating manipulated images of their likeness.

The hearing showed a gap between company claims and real-world results. Menin’s demonstration took less than a minute. The image was generated on a standard version of the product, not a special mode. This raises questions about how effective the company’s current controls actually are.

Kim’s testimony adds a separate layer of concern. He described a workplace where safety staff were few, safety warnings were dismissed, and the line on acceptable content was blurry. His account is part of a lawsuit, so it is contested. But his description of the internal culture matches the public demonstration of the image generation.

The council is now weighing next steps. The company agreed to appear after a subpoena threat. Whether it will face further consequences depends on the outcome of the pending bills and any additional hearings.

For now, the record shows a company representative saying one thing under oath while a live demo showed the opposite. That contradiction is now part of the official hearing record.