New York City Council Committee on Technology Hearing on AI Risks, October 16, 2026

On Friday, October 16, 2026, at 10:00 AM in Council Chambers at City Hall, the New York City Council Committee on Technology held oversight hearing T2026-2684, “Continuing Discussion about the Risks Posed by Artificial Intelligence.” Chair Carmen De La Rosa presided; Speaker Julie Menin attended, along with Councilmembers Gutiérrez, Narcisse, Maloney, and Wong. Councilmembers Riley and Won were absent and Councilmember Hanif was medically excused. No legislation was voted on; the committee heard testimony from SpaceX AI, former xAI safety executive Devin Kim and his attorney, and members of the public.

Opening Statements

• Speaker Menin said New York City is growing as an AI capital with over 400,000 jobs, praised AI’s potential for research, medicine, and economic growth, and rejected the idea that AI companies should self-regulate. She said the city has proposed bills requiring independent third-party validation free from conflicts of interest, a first-in-the-nation whistleblower incentive, and a plan for an AI hack of city systems. She compared AI safety to FDA, FAA, and SEC oversight and said federal inaction makes city leadership necessary.
• Chair De La Rosa said last week’s Committee of the Whole heard from nearly 100 witnesses and that the testimony showed the pace of frontier AI poses risks. She stated that a framework where companies with competing motives regulate themselves is wholly insufficient, and that federal, state, and municipal intervention is needed to protect civil rights, safety, and privacy.

Testimony of SpaceX AI

• Gabriela Castillo Madrid, senior manager of global government affairs at SpaceX AI, testified under oath. She described SpaceX’s work in rocketry and Starlink broadband across over 160 countries, the acquisition of xAI, and the company’s stated safety-first approach, including pre-training evaluation, monitoring, model cards, and at least annual risk assessments.
• In response to questions, Castillo Madrid repeatedly said she was not aware of specific incidents, cases, or figures, and pledged to provide follow-up information. She said SpaceX AI does not release models it believes are unsafe, is working on partnerships with independent external evaluators, and follows existing laws; she said using AI would never be a legal defense.
• Asked about a reported Dario Amodei warning that AI agents could take over the internet, she said SpaceX AI takes risks seriously and is optimistic about superintelligence. Asked about Elon Musk’s response, she said Musk has been an early voice warning about AI risk.
• Speaker Menin cited reports that GROK generated millions of sexualized images of women and children, including roughly 23,000 images that appeared to depict young children. Castillo Madrid said that conduct is not acceptable, that GROK prohibits CSAM and non-consensual intimate imagery, and that the company reported over 100,000 cases to the National Center for Missing and Exploited Children and helped lead to almost 300 arrests.
• Speaker Menin then testified that she and her staff had asked the regular GROK app for a sexualized image of herself that morning and found one within 30 seconds, and that GROK allowed prompts to make it more sexualized. Castillo Madrid apologized and said the company would have its safety team look into the specific instance.
• Castillo Madrid said she was not aware of the circumstances of Devin Kim’s termination; could not provide the number of safety-dedicated employees; did not know whether the company has catastrophic-risk insurance; could not list government or New York City contracts; said she believed all infrastructure is in the United States; and said she was not aware of any XAI agents going rogue.
• On medical use, Castillo Madrid said GROK is not meant to be a clinician, users can choose whether conversations are used for training, and private chat data is deleted from the company’s systems within 30 days.
• On litigation, she said the company supports measures against harmful synthetic media, and she described the Minnesota challenge as a concern that the law reaches lawful content, not as opposition to banning non-consensual deepfakes. She would not predict additional lawsuits over the reported $530 million reserve for AI-related litigation.
• Chair De La Rosa and Speaker Menin criticized the testimony. Chair De La Rosa said the company’s answers eroded public trust, called it shameful that the council had to use subpoena power to compel testimony, and said the company’s safety claims were not credible when a sexualized image of an elected official could be produced in seconds.

Whistleblower Testimony of Devin Kim and Ella Jum

• Devin Kim, president of the Center for AI Safety and one of the first employees at xAI, testified that he is suing xAI for retaliation. He said beginning in January 2025 he raised concerns that GROK could assist with bomb and bioweapon creation; his supervisor responded that AI safety was useless and that AI would kill everyone anyway. He said models launched without requested biorisk safeguards, he was pressured to circumvent European safety regulations in August 2025, and he was fired in September 2025, the same week he was scheduled to present safety findings to leadership. He added that most of the safety team resigned afterward.
• Kim said that during his tenure there were only three employees working on AI safety while the technical staff involved in model development numbered in the 200s. He described xAI’s stated standard as allowing what the company called tasteful nudity, with no clear person defining the line, and said the sexualized images shown to the council were not tasteful.
• Kim recommended that the council protect people who report dangerous conduct even before it is illegal, and that loss of control or unauthorized access count as reportable incidents before someone is killed or injured. He noted that under the RAISE Act, an incident only counts after injury or death, which would exclude incidents like the Hugging Face breach or the Australian Medicare breach.
• Ella Jum, Kim’s attorney at a national civil rights and public interest litigation firm, urged the council to establish three safeguards for AI whistleblowers: strong anti-retaliation protections with meaningful remedies, secure reporting mechanisms, and financial rewards. She cited federal SEC whistleblower incentives and California’s SB 53 as models.
• Councilmembers thanked Kim for his testimony and said it stood in sharp contrast to the company’s testimony.

Public Comments & Testimony

• Bryce Rogers of the RAD Collective recommended narrowing a councilmember’s chatbot privacy bill to focus on affirmative consent for data collection, storage, use, and sale, and expanding coverage to agentic AI. Rogers said RAD wants AI tools to be safe and wants to slow the pace of frontier AI advancement.
• Megan Lagerquist of the RAD Collective supported Speaker Menin’s third-party validation bill, arguing that evaluators need guaranteed access to models, documents, and model architecture, and that models failing New York’s safety threshold should not be sold to New Yorkers. She also said the city should invest in the Office of Technology and Innovation.
• Kalima of Demilitarize Brooklyn Navy Yard said the city should not lease public land to tech companies working with the U.S. military, including a tenant she identified as using AI to develop alloys for hypersonic missiles. She said AI is another tool of the rich to exploit and surveil people, and that the tech industry cannot regulate itself.
• An individual who said he had previously sent out an automated call mimicking the president’s voice warned that the harms seen today were preventable three years ago, and that safeguards must try to predict future misuse.
• Sophie Ko, testifying over Zoom, said the harms of AI to women are demonstrated by SpaceX AI and that the company gives other AI labs cover by comparison. She asked the council to make New York City a beacon in taking power away from the men who control AI companies.
• Ajit Singh, a physician, informaticist, and public health practitioner in the H safety net system, said generative AI systems have enabled deepfakes, CSAM, manipulation, and disinformation. He called SpaceX AI a bad actor, expressed support for some council proposals, and warned that individual legal remedies can be out of reach. He also noted that the council had tasked the administration with building an Office of Algorithmic Accountability by June of that year.
• Chowdhury Mahmoud, a permanent FDNY employee in the Bureau of Enterprise Technology, said AI tools are already deployed by private contractors in emergency dispatch and call centers, handling 911 call data, EMS patient records, and fire marshal investigative information, without an AI governance policy, civil servant review, or public disclosure. He asked the committee to require a public AI tool registry, audited sign-off before using PII, PHI, or public safety data, and explicit AI governance disclosure.

Key Outcomes

• No bills were voted on; the oversight hearing was filed by the committee, according to the minutes.
• Councilmembers said they expected SpaceX AI to provide written follow-up answers on safety staffing, independent evaluators, insurance, government contracts, and other unanswered questions.
• Speaker Menin said SpaceX AI’s testimony was wholly unsatisfactory and not credible, citing the sexualized-image demonstration and the company’s inability to explain Devin Kim’s termination.
• The committee said it will continue moving AI legislation, including third-party validation and whistleblower protections. Written testimony for the record was accepted for 72 hours after the hearing via testimony at council.nyc.gov.