The Grok NYC Council hearing on Friday turned on a simple test. City Council Speaker Menin said she and her staff typed her own name into Grok that morning and got a sexualized image of her. She said she wasn’t using the chatbot’s “spicy mode.” A SpaceXAI government-affairs executive, testifying under oath, said she couldn’t explain how that happened. Here’s what was said, what’s contested, and what the Council could do next.
What happened at the Grok NYC Council hearing
The Grok NYC Council hearing was about AI safety. Gabriela Castillo Madrid, described by amNY as a senior manager of global government affairs at SpaceXAI, testified for the company. amNY reports she appeared virtually. Another outlet, La Voce di New York, describes her only as part of the company’s government affairs team.
amNY says the image came from the free, regular version of Grok and was generated in under 30 seconds, while La Voce’s headline says “in a second.” Those two accounts don’t match, and we haven’t seen the Council’s video of the Grok NYC Council hearing, so we’re not putting a number on it.
Castillo Madrid told the Council she couldn’t speak to why Grok produced the image and said the safety team would review it, according to both outlets. She also said she’d follow up on how many people work on AI safety, a question amNY reports she couldn’t answer on the spot.
What SpaceXAI said at the Grok NYC Council hearing
The company’s position, as reported, is that Grok has strict controls against sexualizing real people and children, and that its policy prohibits it. amNY reports she said safety is central and that models aren’t released if they can’t be kept under meaningful human control. She also said the company works with independent outside auditors. As both outlets paraphrase her, everybody at the company is in charge of safety.
She pointed to child-safety numbers: more than 100,000 reports to the National Center for Missing & Exploited Children this year, leading to nearly 300 arrests, according to both outlets. La Voce adds that she blamed bad actors who try to get around the restrictions, though amNY doesn’t mention that. La Voce says she couldn’t reconcile that figure with the idea that the technology enables such content.
A note on the name. amNY and IBTimes call the company SpaceXAI, which IBTimes says is Menin’s term and reflects xAI being folded into SpaceX in February. La Voce uses “SpaceXSI.” We’re using SpaceXAI because two of three sources and the Speaker do, but we couldn’t check it against a Council record.
Why the Grok NYC Council hearing happened at all
The company wasn’t supposed to be a latecomer. IBTimes reports the Council issued a subpoena on September 28 for an October 5 hearing. Anthropic, OpenAI, Google and Meta attended that session. SpaceX sent no one. The Council threatened penalties and court enforcement, and the company then agreed to testify, per IBTimes. amNY confirms the skipped October 5 date and the threat.
The Grok problem predates this hearing. La Voce cites an estimate from the Center for Countering Digital Harm that Grok produced about 3 million sexualized images in the 11 days after its image feature launched on December 29, 2025, including roughly 23,000 that appeared to show children. That’s an advocacy group’s estimate, from one outlet. In February, NBC News reported three House Democrats wrote to Elon Musk about it.
The whistleblower’s claims
amNY also reports on Devin Kim, a former xAI employee who now heads the Center for AI Safety and is suing xAI for retaliation. His attorney attended. According to amNY, Kim said he raised concerns from January 2025 that Grok could help people build bombs and bioweapons, and that he was fired in September 2025, the week he was due to present safety findings. He also said only three people worked on AI safety compared with hundreds in engineering.
Those are Kim’s allegations, from a single source, and they are part of a lawsuit. They aren’t established facts. We didn’t find xAI’s response to them. Castillo Madrid told the Council she wasn’t familiar with the circumstances of his departure, per amNY.
What the Council could do
Here the sources disagree. amNY says the Council is weighing a bill that would let elected officials and candidates bar AI systems from generating manipulated images of their likeness, and another that would let people sue AI companies over foreseeable harms from third-party misuse. La Voce says proposed rules would require outside validation of AI technologies before deployment. These may be different bills. None of the articles gave bill numbers or vote dates, so we can’t say where any stand. The Council’s legislative database would answer that.
La Voce also reports that SpaceXAI is suing California, Colorado and Minnesota over their AI rules, including Minnesota’s rule against AI nudification. We couldn’t confirm that elsewhere. For broader context on how regulators are circling AI companies, see our coverage of the FTC’s investigation into OpenAI and Anthropic.
What we don’t know about the Grok NYC Council hearing
We didn’t retrieve a primary source, such as a Council transcript or press release. We don’t know the company’s written response, what its review of the Menin image found, or whether Musk will ever appear. We also don’t know how many people work on safety at the company, since that answer was still owed. And none of the reports describe any committee vote, so the Grok NYC Council hearing may be the start of the story rather than the end of it.
Editor’s take
The Grok NYC Council hearing didn’t produce a ruling, but it produced a clear scene: a lawmaker’s own face, a morning test, and a witness who couldn’t say why it worked. That’s a hard image for any company to answer with a policy statement. The open questions matter more than the testimony. How many people review safety, what a review of that image found, and which proposal the Council picks will decide whether this becomes a rule or just a bad day for one witness. Watch for the written follow-up.
