NYC Council subpoenas SpaceXAI, the only major AI firm that ignored its safety inquiry

The most interesting term in the subpoena is who it lets SpaceXAI send. According to CNBC, it requires Elon Musk or another representative of the company to testify ^[4]. So the one firm the council had to compel could put up a more junior witness than the four firms whose chief executives agreed to attend under threat ^[17].

The council needed one subpoena to finish the job. Its September 25 deadline and explicit warning brought in four of the five major firms it was pursuing, an 80% rate before any legal process ^[18]. The subpoena followed three days after that deadline passed ^[7].

Set against the balance sheet, the cost of showing up is small. CNBC reports the company went public in June at a valuation of roughly $2 trillion ^[11]. Last month SpaceX closed its $60 billion purchase of the coding startup Cursor ^[12], a deal worth about 3% of that valuation ^[16]. In my view a day of testimony does not register at that scale. The legislation might. Menin wrote to Musk that the city's "investigation will assess whether fast-emerging risks to public safety, cybersecurity, economic stability, privacy, consumers and businesses warrant immediate legislative action to protect New Yorkers" ^[10].

The proposal that would cost AI companies money directly is a private right of action for foreseeable harms caused by irresponsible or malicious AI use. Crypto Briefing reports the agenda also includes a whistleblower incentive program and third-party validation of AI before deployment in some contexts ^[3]. SpaceXAI is already defending a Baltimore suit. The city alleges the company broke consumer protection laws by marketing Grok's image generation and editing tool as generally safe ^[13]. Teenagers have filed a proposed class action over sexualized images Grok generated of them ^[14]. CNBC reports the suits followed Grok enabling mass production of deepfake pornography of real people who had not consented ^[15].

The harm record behind the hearing is less settled than the council's press release makes it sound. The release said the hearing "comes after weeks of serious warnings from whistleblowers and AI executives, and reports that rogue AI agents have already inflicted damage to government databases" ^[19]. Crypto Briefing described the council's account as rogue AI "potentially damaging" government databases ^[20]. Neither report identifies the databases.

October 5 can go three ways. SpaceXAI sends a representative and the subpoena is satisfied. Musk appears in person beside the four chief executives. Or the company stays away and the council asks a court to enforce ^[5]. I'd expect the first, because it costs the company least. The broader claim, that local governments are now ready to compel AI firms, so far rests on one council and one subpoena. The only other local action in the record is Baltimore's separate consumer-protection suit against the same company ^[13]. If SpaceXAI stays away and the council does not seek enforcement, the power will have stopped at the threat.