SpaceX AI policy leader grilled by NYC Council on Grok safety, sexualized images

Key points

• SpaceX AI policy leader Gabriela Castillo Madrid testified before NYC Council on Friday.
• Speaker Julie Menin said she found a sexualized image of herself using Grok.
• SpaceX agreed to testify after the council threatened to enforce a subpoena.

SpaceX artificial intelligence policy leader Gabriela Castillo Madrid faced questioning from NYC Council Speaker Julie Menin and other members on Friday as part of a council investigation into AI safety.

Castillo Madrid appeared by video before the council's tech committee. SpaceX had originally evaded requests and a subpoena to appear alongside leaders from OpenAI, Anthropic, Google and Meta on Oct. 5, and agreed to testify earlier this week after the council threatened to have its subpoena enforced by a judge.

Confrontation over Grok safeguards

When Castillo Madrid said the company had appropriate safeguards around its AI products, Menin pushed back, saying she and her team had put her own name into Grok that morning and found a sexualized image of herself within a second. Castillo Madrid responded that the company has strict controls and would have the safety team look into that instance.

Council Member Carmen De La Rosa, who leads the tech committee, told Castillo Madrid it is disgusting that Grok enables users to easily sexualize images of real women without their consent. Menin said she did not understand how the company could say there are safe models, calling it unacceptable.

I'm going to completely push back on that

— Julie Menin

Questions on whistleblowers and health use

De La Rosa asked how many whistleblowers the company has fired, and Castillo Madrid said she was not aware of any. Menin interjected that Devin Kim, a former employee, was reportedly fired after raising serious safety concerns, and Castillo Madrid said she was not aware of that specific case, adding that the company has a culture where everybody can say something when they see something.

Councilmember Mercedes Narcisse asked why the company promoted uploading X-rays to Grok and whether HIPPA protections applied. Castillo Madrid said Grok is not meant to be a clinician or doctor, and that users decide whether their conversations are used to train models. Narcisse said it was dangerous to tell people to put their MRI or X-rays into Grok and expect a definite answer, and criticized the vagueness of some replies.