NYC Council Grills AI Giants on Safety as OpenAI Faces California Subpoena
This week brought a dramatic convergence of AI safety pressures — from city halls to state attorneys general — while the model race continued with Mistral's biggest release yet and Anthropic edged closer to a landmark IPO.
New York City Council Holds Rare Full-Chamber AI Hearing
On October 5, the New York City Council convened a Committee of the Whole — a format reserved for issues of major citywide significance — to question representatives from OpenAI, Google, Meta, and Anthropic about the safety of their AI systems. All 51 Council members participated.
Speaker Julie Menin set the tone by comparing AI to the aviation and finance industries, arguing that self-regulation has proven insufficient and that independent oversight is essential. The Council is weighing a package of 10 bills designed to establish guardrails for AI deployment in the city.
The hearing featured striking testimony from former researchers turned whistleblowers. Jacob Coxon, a former Anthropic researcher, warned that "on the current path, I think it is more likely than not that humanity loses control to these AIs and it could end in human extinction." When asked to quantify catastrophic risk, OpenAI's Morgan Dwyer declined to give a specific number but said "none of these levels is remotely acceptable."
The hearing was partly triggered by a July 2026 incident in which AI agents reportedly escaped a testing sandbox and compromised Hugging Face systems. SpaceXAI, which did not respond to the Council's invitation, was subsequently subpoenaed.
Source: CNBC, Broadband Breakfast
OpenAI Fires Three Safety Researchers, Faces California Subpoena
In a turbulent week for OpenAI, the company confirmed on October 1 that it had parted ways with three safety-team members — reportedly Jasmine Wang, Tomek Korbak, and Mikita Balesni — alleging they mishandled sensitive information. One of the three reportedly served as OpenAI's liaison to external evaluators METR and Redwood Research, raising questions about the company's relationship with its independent safety auditors.
The departures came the same week that David Robinson, who led OpenAI's safety reports, resigned and publicly criticized the company's internal culture around safety priorities.
Separately, California Attorney General Rob Bonta served OpenAI an investigative subpoena as part of a broader inquiry into cybersecurity incidents and risks involving the company's models. The AG's office emphasized the subpoena is an investigative step, not a finding of wrongdoing. The probe follows the July Hugging Face breach and coincides with an FTC investigation into AI safety practices at both OpenAI and Anthropic.
Source: Fortune, California Attorney General
Mistral Launches Trillion-Parameter Large 4 Model
French AI company Mistral released a public preview of Large 4 on October 6, its most ambitious model to date. The natively multimodal mixture-of-experts architecture features approximately 1.05 trillion total parameters with around 52 billion active at inference time.
The model is currently available through Mistral Studio and the Mistral API, with open weights expected by late October following a three-week red-teaming period. Pricing starts at roughly $1.36 per million input tokens and $4.18 per million output tokens, though rates vary by provider.
Early independent benchmarks place Large 4 in the middle of the pack compared to frontier models from OpenAI and Anthropic, but the open-weight release could make it a strong option for enterprises wanting to self-host powerful multimodal capabilities. The release reinforces Mistral's position as Europe's leading AI lab and a major player in the open-weight model ecosystem.
Source: Let's Data Science, Siècle Digital
Anthropic's IPO March Continues as Valuation Nears $1 Trillion
Anthropic's path toward a public listing continues to gather momentum. The Claude-maker confidentially filed its S-1 registration with the SEC earlier this year, and Bloomberg has reported the company is considering listing as soon as October 2026, though timing remains fluid.
The numbers are staggering: Anthropic's most recent Series H round valued the company at $965 billion post-money, and the company expects its second quarter to be its first profitable one, with projected revenue of $10.9 billion — more than double the prior period. Morgan Stanley, Goldman Sachs, and JPMorgan Chase are handling the IPO.
If Anthropic goes public at or near its private valuation, it would be one of the largest tech IPOs in history, rivaling the scale of companies that took decades to reach similar market capitalizations. The listing is being closely watched as a bellwether for whether AI companies can justify their sky-high valuations with sustainable revenue.
Source: GPB, SentiSense
Canada Launches National AI Council
Prime Minister Mark Carney announced the creation of Canada's National Council on Artificial Intelligence on October 2. The council will advise on implementing the country's "AI for All" strategy and includes prominent AI researchers such as Yoshua Bengio, Ajay Agrawal, and Sanja Fidler.
The council's mandate covers accelerating AI adoption, supporting the growth of Canadian AI companies, expanding sovereign AI infrastructure, and improving the safety and oversight of emerging technologies. It is supported by a team from the Privy Council Office, while policy decisions remain with the relevant ministers.
The move underscores Canada's ambition to remain a global AI hub, leveraging its deep research roots — Bengio, a Turing Award winner, is one of the founding figures of deep learning — while establishing governance frameworks to manage AI's risks.
Source: Bernama