New York City Council begins AI safety oversight hearing
New York City Council began an artificial-intelligence oversight hearing involving Anthropic, OpenAI, Google and Meta while considering 10 bills. Council Speaker Julie Menin said the city wanted answers about safety measures, including a proposed shutdown mechanism. SpaceX declined to attend, and Menin said the council was pursuing legal action over its response to a subpoena.
Artificial Intelligence··Night
Council hears the companies
New York City Council began its artificial-intelligence oversight hearing on October 5, with representatives of Anthropic, OpenAI, Google and Meta taking part. The hearing brings technology-company representatives before the council to discuss AI safety. Whistleblowers are also appearing, presenting concerns about the development and oversight of increasingly capable systems alongside the companies’ accounts of their own work.[1], [2]
Council Speaker Julie Menin said the city needed to act because the federal government had allowed AI companies to regulate themselves. The council is considering 10 AI-related bills. Menin’s remarks describe the city’s reason for convening the hearing and pursuing legislation; the bills are under consideration, rather than measures already enacted through this hearing.[1]
A shutdown proposal remains under discussion
One question concerns a proposed mechanism for shutting down an AI system. Menin said the idea had also been discussed in New York State and California and appeared in city legislation being considered. Council members wanted to ask companies how such a mechanism would operate and whether it would be effective. Its technical workings and effectiveness remain questions for the hearing.[1]
Former Anthropic employee Jacob Coxon was among the witnesses. He raised concerns about AI systems approaching the ability to perform research previously carried out by people and contribute to their own improvement. His testimony describes his assessment of emerging capabilities and risks. It does not establish that the systems have already reached every capability he discussed or that a particular future outcome is certain.[1]
SpaceX declines attendance
SpaceX said it would not attend. According to Menin, its response to a subpoena was a letter expressing a wish to work with the city while declining to appear at the hearing. Menin said the council was pursuing legal action over that response. The reported action concerns the company’s refusal to attend; no final court decision on that dispute is described.[1]
Before the hearing, OpenAI sent a letter describing testing for dangerous capabilities, stronger security and monitoring for failures in safeguards. Meta also stated that safety was central to its work. Those statements set out the companies’ descriptions of their own safety efforts. They accompany the council’s questions about regulation and proposed mechanisms rather than resolving those questions in advance.[1]
