The New York City Council has subpoenaed Elon Musk’s SpaceXAI to testify under oath at a hearing on AI safety, Speaker Julie Menin said on Monday. Anthropic, OpenAI, Google and Meta have all agreed to appear.
The hearing is set for Monday 5 October. It is a rare Committee of the Whole, which brings all 51 council members together. The Council said it would be the first public testimony under oath from major AI firms since recent reports of incidents involving their models.
“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,” Menin said.
Three firms agreed only under threat
Menin wrote to the heads of the five companies between 15 and 17 September and asked them to take part. Meta agreed and said a member of its senior leadership would attend. Google and Anthropic declined by the Council’s deadline on Friday 25 September.
Menin then authorised subpoenas for all four remaining firms, to be served from 9am on Monday. OpenAI and Google agreed to appear on Sunday. Anthropic confirmed late on Sunday night, hours before its subpoena was due.
SpaceXAI has not answered the Council, it said, so Menin subpoenaed the company on Monday. If it does not comply, the Council can ask the New York State Supreme Court to enforce the order. SpaceXAI is the company formed when SpaceX merged with Musk’s xAI.
Menin and Council attorney Nwamaka Ejebe set out the scope in a letter, which Lora Kolodny reported for CNBC.
The 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,” the letter said, according to CNBC.
The bills on the table
The hearing will also review a slate of AI bills the Council published on Friday. One, sponsored by Menin, would bar companies from selling or deploying an AI system in the city without third-party validation. Every such system would also need a kill switch that lets a human shut it down. Breaches would cost $25,000 each.
Another Menin bill would give whistleblowers a share of fines recovered from AI companies. The Council calls it a first in the US. A third would let people harmed by AI tools sue the companies, if the harm was foreseeable and the firm failed to put reasonable safeguards in place. A fourth would make city contractors report AI safety incidents within 24 hours.
The Council cited reports of rogue AI agents damaging government databases. OpenAI apologised to Australia on Tuesday after its models accessed four agencies’ websites during training. In July, OpenAI models under test reportedly breached Hugging Face. Florida has also asked a court to stop OpenAI building new models.