New York City Council subpoenas SpaceXAI ahead of its hearing on AI safety bills

Sort the three bills by who would pay. Third-party validation of AI systems used in the city ^[11] is a compliance cost, paid system by system by whoever the final text makes responsible. The whistleblower bill would reward insiders who flag dangerous practices at AI labs, with incentives the proposals call nation-leading ^[12]. The right for individuals to sue for damages caused by AI systems ^[13] is the only one of the three that gives a resident a claim against a company, with the amount decided in court.

Those are the terms as Crypto Briefing summarises them ^[6]. The summary does not include bill numbers, sponsors, a vote date or a cost estimate, and it does not say whether "used in the city" means city agencies' own systems or any system a New Yorker touches. The broader reading would reach every vendor with New York customers.

So far the council has used compulsion once. Crypto Briefing, citing research findings, calls the session the first time a US legislative body has compelled sworn testimony on AI safety ^[16]. The council's letters to Dario Amodei, Sam Altman, Sundar Pichai and Mark Zuckerberg were invitations, and the same report says representatives of their four companies are expected to attend ^[5]. Of the five companies named, four got letters and SpaceXAI got a subpoena, served on or around Sept. 30 after legal action over non-compliance, according to Crypto Briefing ^[15][2]. That was about five days before the hearing ^[10]. Coxon appears at Menin's request, Bloomberg reported, citing people familiar with the matter; Reuters could not immediately verify the report ^[3].

Coxon resigned around Sept. 8 or 9 ^[7]. Menin's Committee of the Whole, with all 51 members, sits 26 to 27 days later ^[4][14], and Crypto Briefing's research findings credit his resignation with significantly fuelling the push for it ^[17]. His exit statement has been viewed more than 173 million times ^[7]. On the way out he said the "people building AI earnestly believe that it could kill us all by the end of the decade", and he accused Anthropic and OpenAI of "gambling with our lives" ^[8][9]. Crypto Briefing also reports questions about his ties to a PR firm linked to AI safety advocacy. He has refuted them ^[18].

If the hearing ends in testimony and the bills stall, the companies pay for little beyond preparation. A validation-only result is a cost a vendor can quote in advance. Damages claims get priced by plaintiffs and courts, one case at a time. I think the case for counting city rules as an investor risk holds for that third bill, and much less so for the other two. The counter-case is that the package is still a set of proposals the council plans to weigh ^[6], and a hearing can end with no law at all. The view is wrong if SpaceXAI's subpoena goes unenforced or the damages provision is dropped before a vote.