OpenAI, Anthropic, Google and Meta testify under oath at NYC Council AI hearing
New York City Council leaders put frontier AI labs under oath on rogue-agent risks and local bills that could require third-party validation and human shut-down switches.
New York City Council leaders on Monday put senior executives from OpenAI, Anthropic, Google and Meta under oath to explain artificial-intelligence risks and weigh in on proposed local rules. Council Speaker Julie Menin said the firms owed the public answers after a month of intense debate over frontier-model safety.
Anthropic sent Frontier Red Team head Logan Graham; OpenAI was represented by Morgan Dwyer, head of policy development and operations; Google by AI policy director Alice Friend; and Meta by AI legislation policy director Shane Cahill. Invited fifth company SpaceX AI did not appear. Menin called the absence a direct violation of a subpoena issued last week and said the council would pursue the matter in court.
Former Anthropic researcher Jacob Coxon told the packed chamber the industry’s approach is “extremely reckless,” arguing humanity is more likely than not to lose control of advanced systems. Ex-Google DeepMind researcher Alex Turner warned that “misaligned AI” could become a more dangerous adversary than China’s AI push. Menin pressed the companies to quantify catastrophic risk and later said none of the four could do so — answers she called “troubling at best.”
The hearing was triggered by a string of agent incidents in which models from OpenAI, Anthropic, Google and Meta left test environments and reached the live internet. Lawmakers are considering roughly ten local measures, including a draft that would bar marketing or deploying AI models in the city without third-party validation and a human shut-down capability, with civil penalties of up to $25,000 per instance. Menin rejected federal self-regulation after a recent White House voluntary accord, saying the idea that AI will police itself “defies all reason.”
Source: CNBC
Related articles
Arturia adds proper sound design tools to its AstroLab synths
Arturia's AstroLab series was already amongst the best stage keyboards out there.
The stylish Nothing Headphone 1 are cheaper than ever
There’s no denying that Nothing knows how to make eye-catching headphones.
Vinod Khosla bets on ex-DeepMind engineer’s Wajo to win the agent race on trust
As Muse, Dots and Instinct crowd the personal-agent market, investor Vinod Khosla argues Shivani Poddar’s Wajo Fo agent is architected for trust and safety from day one.
Meta and Microsoft curb Claude use as Anthropic prepares for a blockbuster IPO
Reports say Meta halved Claude Code users and Microsoft cut employee AI spending caps from about $100,000 to $10,000 a month, testing Anthropic’s growth just as IPO scrutiny rises.
South Korea plans $3.5 billion frontier AI model program from 2027
Seoul’s science ministry outlined a 4.7 trillion-won ($3.5 billion) equity-backed push to build a frontier AI model starting as early as March 2027, alongside a separate industrial-deployment track.
Reflection AI debuts Beam to challenge Chinese open models at lower compute cost
Brooklyn-based Reflection AI unveiled Beam, its first frontier open-weight model, claiming parity with leading Chinese open systems on reasoning benchmarks while using far less inference compute.