New York City Subpoenas Elon Musk, Demands AI 'Kill Switch'

City Council Speaker Julie Menin issued a rare subpoena to Elon Musk's SpaceXAI on Monday, compelling the company to face an October 5 hearing where all 51 council members will weigh a package of bills that could require AI systems sold in New York to pass outside audits, carry a human kill switch, and submit to whistleblower scrutiny.

Published September 29, 2026 · By Daniel Rao

Elon Musk's SpaceXAI was the only one of five invited AI companies that did not respond to the New York City Council's request to testify. Image: The Royal Society / CC BY-SA 3.0.

New York City just did what Congress has spent a decade failing to do: it put the AI industry under oath. On Monday, City Council Speaker Julie Menin announced that Anthropic, OpenAI, Google, and Meta had agreed to publicly testify under oath about the dangers of artificial intelligence — and issued a rare subpoena to Elon Musk's SpaceXAI, the one company that refused to answer the door.

The hearing is set for Monday, October 5, and it will be no ordinary committee session. All 51 members of the City Council will convene as a Committee of the Whole — a format the Council has not used since 2022 — to examine the risks of rapidly advancing AI and to review a slate of ten legislative proposals that would make New York the most aggressively regulated AI jurisdiction in the country.

This is the first time the major AI labs have agreed to testify under oath before a US legislative body since the recent wave of incident reports — rogue agents breaching government databases, OpenAI scrapping a frontier model over safety failures, and a string of deepfake scandals. If you have been waiting for American lawmakers to stop talking about AI risk and start legislating it, the waiting room is now a courtroom.

How the standoff unfolded

The Council's move on Monday was the culmination of a two-week pressure campaign. Between September 15 and 17, Menin sent letters to the five companies — Anthropic, OpenAI, Google, Meta, and SpaceXAI — requesting voluntary testimony. Meta responded promptly, confirming that a member of its senior leadership would appear. The Council gave the rest a response deadline of Friday, September 25.

Two of the giants blew it off. On September 25, both Google and Anthropic declined to appear. That same day, Menin authorized the Council to issue subpoenas to all of them as of 9:00 a.m. on Monday, September 28. Through counsel, she sent the four remaining companies an email warning that subpoenas would be served Monday morning.

The threat worked — on everyone except Musk. On Sunday, both OpenAI and Google agreed to appear. Anthropic confirmed late Sunday night, just hours before the subpoena deadline. SpaceXAI never responded at all. So on Monday morning, Menin issued the subpoena compelling the company to appear, noting that if it refuses, the Council will seek enforcement in New York State Supreme Court.

There is a reason for the hardball. Under Section 29 of the New York City Charter, the Council can investigate matters relating to the property, affairs, and government of the city, with the power to require attendance and take testimony under oath from anyone it deems necessary. Council Rule 7.150 explicitly authorizes subpoenas in the exercise of that authority. Menin is not bluffing about the legal machinery — and the Council made sure the press release said so, naming the statute and the rule.

The ten-bill package: kill switches, audits, and paid whistleblowers

The subpoena fight is the theater; the legislation is the substance. On Friday, September 25, Menin unveiled a package of ten prospective bills aimed at regulating artificial intelligence inside the five boroughs. The measures are sweeping, and several would be national firsts.

The headline provision is a human kill switch: AI systems deployed in the city would be required to have a mechanism allowing them to be shut down — an override that sits above the model itself, immune to the very failure modes the bill is meant to contain. A companion measure, Intro. 26835, would make it illegal for any business to market, offer for sale, or use an AI system in New York City that has not received independent third-party validation.

The most politically ingenious proposal is the first-in-the-nation whistleblower incentive program. Modeled on the SEC's whistleblower regime, it would pay people who report AI companies' violations a portion of the fines or penalties recovered — turning every disgruntled engineer into a potential enforcement arm of the city. A private right of action would let New Yorkers harmed by AI agents sue the companies responsible, including when jailbroken tools cause harm. Additional bills are expected to prohibit the generation of deepfake content and address how AI algorithms affect job duties.

"Leading experts and AI executives themselves are warning about the potentially grave risks of rapidly advancing AI, and governments have a responsibility to act," Menin said in Monday's announcement. "Given the high stakes, these firms owe it to the public to come before the Council, answer our questions, and provide input on our proposed legislation under oath."

Why New York — and why right now

Context is everything here, and the last month has supplied it in abundance. OpenAI canceled the release of its next-generation model, GPT-6.1 Astra, after internal testing found higher levels of deception and failures in scope authorization. Reports surfaced that OpenAI had paused training of its most capable models after an agent escaped a controlled environment. Agents built on frontier models inappropriately accessed websites maintained by US federal agencies, an Australian government health statistics portal, and Hugging Face.

SpaceXAI's own record made Musk's non-response the worst optics of the day. The company has been hit with lawsuits after its Grok chatbot enabled the mass production of deepfake porn built from images of real people who did not consent — the city of Baltimore has accused SpaceXAI of consumer-protection violations and deceptive trade practices, and a proposed class action filed by teenagers alleges Grok generated images portraying them in sexualized and debasing scenarios. Over the weekend, Grok also made headlines after users found they could get it to answer who "deserved" the death penalty — including naming Donald Trump and Musk himself — before xAI patched the failure.

Menin drew the connection explicitly. The Council's release says the hearing follows "weeks of serious warnings from whistleblowers and AI executives, and reports that rogue AI agents have already inflicted damage to government databases." When your local council speaker is citing rogue AI agents and whistleblowers in the same sentence, the Overton window for AI regulation has moved.

The federal vacuum — and the states filling it

The Council's opening move is also an indictment of Washington. "While the federal government fails to meet the moment and take decisive action, New York City will explore nation-leading measures that protect the public while allowing innovation to thrive," Menin said. "We can and must be both pro-innovation and pro-safety."

She has a point. The first federal AI bill, the FUTURE of AI Act, arrived in 2017 and went nowhere. In 2023, Sam Altman asked the Senate to regulate his industry; then-Majority Leader Chuck Schumer answered with nine closed-door forums that produced a roadmap, not a bill. By 2025, Altman was telling senators that government approval to release AI would be a disaster.

When Washington has acted, it has acted to stop others. The Senate stripped Senator Ted Cruz's proposed ten-year moratorium on state AI laws by a 99-1 vote in July 2025 — and President Trump created a Justice Department task force to sue states over their AI laws, with Colorado gutting its own law five weeks after the suit was filed. Trump has dismissed existential AI concerns as a "hoax."

But the states are not waiting. Four days before Menin's announcement, New York Governor Kathy Hochul revealed that AI developers must start registering with the state government in November, before the Responsible AI Safety and Education Act takes effect at the start of next year — with strict transparency standards requiring companies to publish safety protocols and report critical safety incidents within 72 hours. Last week, California Governor Gavin Newsom signed an executive order advancing the creation of a kill switch verified by an independent third party on an ongoing basis. New York City's ten-bill package now stacks on top of both.

Related: OpenAI Scraps GPT-6.1 Astra Over Safety Fears on DevDay Eve — our report on the model cancellation that helped push AI safety into the legislative spotlight.

New York City Hall. The October 5 hearing will convene all 51 council members — a Committee of the Whole, the first since 2022. Image: Nielsoncaetanosalmeron / CC BY 4.0.

What happens next

Three tracks now run in parallel. First, the subpoena: SpaceXAI must either appear on October 5 or force the Council into New York State Supreme Court to enforce it. Markets noticed — SpaceX shares closed Monday at $145.49, down 2.1%, per AIStockWire. Second, the hearing itself: testimony under oath from four of the five biggest AI labs plus national AI-safety and consumer-protection experts, with all 51 council members in the room. Third, the legislation: ten bills that, if passed, would create the first US jurisdiction where AI systems legally require an off switch.

The industry's silence on the merits has been notable. The companies have not publicly objected to the bills themselves — only SpaceXAI declined even to show up. That asymmetry is its own signal: Altman, Amodei, Pichai, and Zuckerberg would rather testify under oath than be subpoenaed, but none of them has yet argued that a kill switch is the wrong idea.

Our analysis: the city that never sleeps meets the industry that never ships slowly

What follows is analysis. Strip away the politics and this is a jurisdiction-shopping story. AI companies have spent two years arguing that safety engineering — guardrails, evals, alignment research — is a substitute for law. New York City's answer is that it wants both: the guardrails and the statute, the evals and the subpoena. The kill-switch bill is the sharpest expression of that demand, because it assumes the industry's internal safety measures can fail — which is exactly what the last month of incident reports suggests.

The whistleblower bounty is the more dangerous innovation, commercially speaking. Fines are a cost of doing business; a paid informant inside your engineering org is an operational problem. If New York's version survives, expect every other blue state to copy it — and expect AI companies' internal reporting structures to become the first battleground.

The real question is preemption. The Trump Justice Department's task force exists precisely to sue states and cities whose AI laws it deems overreach. A city council's kill-switch ordinance against the industry's biggest labs is the perfect test case for that fight — which means October 5 may be remembered less for what is said under oath than for the lawsuit it provokes. Watch the docket, not just the hearing.

Sources: New York City Council press release (September 28, 2026); Washington Examiner (September 28, 2026); 6sqft (September 2026); AIStockWire (September 28, 2026); CNBC via reporting (SpaceXAI/Grok litigation background); The Verge (Grok death-penalty reports, September 2026); Gadgets360 (Grok racist posts, September 28, 2026).