NYC Council Hears AI Safety Warnings; OpenAI Secures $30B Funding

Key Takeaways

• NYC Council held AI safety hearing on Oct. 5, 2026, with whistleblowers warning of extinction-level risks from uncontrolled AI development.
• Former Anthropic researcher Jacob Coxon warned AI could lead to human extinction; former OpenAI researcher Daniel Kokotajlo said AIs are becoming superhuman at hacking.
• NYC Council is considering 10 bills including a kill switch, whistleblower protections, and deepfake restrictions on elected officials.
• Mistral AI released Mistral Large 4, which outperforms Chinese models in cybersecurity and ranks among the world's best open models.
• OpenAI's GPT-6 Astra scored 55.0% on Ironclad business tasks versus 41.6% for GPT-5.6 Sol, and was 48% faster.
• OpenAI is in $30B funding talks with UAE funds including MGX, seeking a $1.4 trillion valuation.
• Zscaler CEO Jay Chaudhry called agentic AI the biggest opportunity in the company's history during Investor Day.
• Thomas College received a $3 million federal grant to integrate AI into curriculum and create an AI Specialist in Residence position.
• HEAR protocol achieved 1.61 times batch speedup and 2.23 times reduction in time-to-first-token for AI agents.
• Orgvue launched WebMCP interface for AI workforce design, one of the first software vendors to adopt the WebMCP standard.

NYC Council holds AI safety hearing with key witnesses

The New York City Council held a hearing on artificial intelligence safety on Monday, Oct. 5, 2026. Jacob Coxon testified at the event about AI risks. The hearing included photos showing the proceedings and attendees. The event took place in New York City.

NYC Council AI safety hearing draws photographers

A photographer attended the New York City Council hearing on artificial intelligence safety in New York on Monday, Oct. 5, 2026. The event was documented through official photos. The hearing focused on the safety concerns surrounding AI technology development.

AI whistleblowers urge NYC Council to tighten regulation

AI whistleblowers testified at a New York City Council hearing on Oct. 5, 2026, urging stricter regulation. Former Anthropic researcher Jacob Coxon warned that humanity could lose control of AI, possibly leading to extinction. Former OpenAI researcher Daniel Kokotajlo said AI could become a disaster if development is not slowed. Executives from OpenAI, Anthropic, Google and Meta also testified. Council Speaker Julie Menin criticized the companies and said New York City should lead on AI rules since the federal government has not acted.

NYC City Council debates catastrophic AI risks

The New York City Council held a hearing on AI catastrophic risks on Monday, Oct. 5, 2026. Former Anthropic researcher Jacob Coxon said AI could lead to human extinction. Former OpenAI researcher Daniel Kokotajlo warned that AIs are becoming superhuman at hacking. Executives from Google, OpenAI, Anthropic and Meta defended their safety efforts but frustrated Council Speaker Julie Menin with vague answers. The Council is considering 10 bills including a kill switch, whistleblower protections and rules against deepfakes of elected officials.

NYC lawmakers push nation-leading AI safety rules

New York City is positioning itself as a leader in slowing AI development, according to the Wall Street Journal. On Oct. 5, 2026, the City Council considered several bills to rein in AI. Proposed rules include a city-run kill switch for AI, whistleblower protections and restrictions on AI in public schools. AI executives from OpenAI, Anthropic and Meta said they support guardrails including third-party evaluators when asked about worst-case scenarios.

Mistral CEO says new AI model beats Chinese ones in some areas

Mistral CEO Arthur Mensch said the company's new AI model outperforms Chinese models in cybersecurity. He spoke at the Ai Everything conference in Abu Dhabi on Tuesday. The model will be officially unveiled later that day. Mensch said Europe can compete in the AI race. Mistral is independent from both the US and China.

Mistral AI unveils new AI model aimed at narrowing gap with Chinese competitors

Mistral AI released Mistral Large 4 on Tuesday, October 6. It is the company's newest flagship model. Co-founder Guillaume Lample said it narrows the gap with the best models in the world. The model ranks among the world's best open models. Mistral AI is under pressure to perform in a sector dominated by American and Chinese companies.

Advancing computer use with Ironclad

OpenAI partnered with Ironclad to train AI agents on complex business tasks. GPT-6 Astra was the first frontier model trained on Ironclad tasks. It scored 32% higher than GPT-5.6 Sol and was 48% faster. The work focused on contracting workflows like setting up approvals and legal terms. Ironclad provided software environments for models to practice.

Advancing computer use with Ironclad

OpenAI and Ironclad created 11 research tasks across legal, commercial, and procurement work. These included setting up nondisclosure agreements and procurement approval processes. Each task was evaluated against 8 to 50 criteria. GPT-6 Astra scored 55.0% on these tasks compared to 41.6% for GPT-5.6 Sol. Astra also reduced average time per attempt from 37.0 minutes to 19.2 minutes.

New HEAR Protocol helps AI agents communicate with servers more efficiently

Researchers created HEAR, a protocol that connects AI agent harnesses with inference engines. It standardizes how workflow intent and runtime state are shared. HEAR was tested on four benchmarks under memory-constrained conditions. It achieved a 1.61 times batch speedup and reduced time-to-first-token by 2.23 times. The paper was submitted on October 5, 2026.

OpenAI in $30B Funding Talks with UAE Funds

OpenAI is reportedly in talks with UAE investment funds, including Abu Dhabi-based MGX, for a new $30 billion funding round. MGX could contribute up to $10 billion collectively with other investors. OpenAI is seeking a $1.4 trillion valuation. The fundraising comes as the company faces growing capital needs for AI development. BlackRock, Thrive Capital, and Andreessen Horowitz are also reportedly involved in discussions.

Zscaler Investor Day Focuses on Agentic AI Security

Zscaler held its Investor Day in New York City and reaffirmed its financial guidance for the quarter and full fiscal year. CEO Jay Chaudhry called the agentic AI era the biggest opportunity in the company's history. Zscaler stock rose 3% to $207.10. The company's stock is down 8% year to date, lagging peers like CrowdStrike and Palo Alto Networks. The key question is whether agentic AI demand will show up in Zscaler's reported growth.

Thomas College Receives $3 Million Federal AI Grant

Thomas College in Waterville, Maine received a $3 million grant from the US Department of Education. The grant, called Title III Strengthening Institutions, will help the school integrate artificial intelligence into its curriculum. The funding will create new positions including an AI Specialist in Residence. College officials say the goal is to help students develop practical and ethical AI skills before they graduate.

Maydin Media Expands KAGAPSO Framework with AI Systems

Maydin Media LLC is expanding its proprietary KAGAPSO Business Growth Framework by adding AI-powered revenue systems. The framework, developed by founder Mehmet Çayın over 25 years, connects marketing, sales, CRM automation, and AI into one unified system. It has been applied across more than 300 businesses. In 2026, KAGAPSO won a Gold Winner at the Hermes Creative Awards. Çayne also earned the FCIM Fellowship from the Chartered Institute of Marketing in the UK.

New Research Introduces Grounded Skill-Following Framework

A new research paper introduces Grounded Skill-Following, a method that trains AI agents to follow expert-authored skills across multiple phases. The approach uses skill contracts that combine instructions with a runtime specifying required phases and actions. The framework introduces Verified Progress Credit to reward agents at each milestone. It also measures procedural compliance using Protocol Completion Rate. The Qwen3.5-4B model achieved 99.27% PCR on Math and 99.96% on Search tasks.

Orgvue launches WebMCP interface for AI workforce design

Orgvue announced its new Web Model Context Protocol (WebMCP) interface that lets trusted AI agents work directly inside its platform. The tool helps users make faster and more defensible workforce decisions while keeping AI actions visible and auditable. No organizational data changes without explicit human review and validation. Orgvue is one of the first software vendors to adopt the WebMCP standard. An early access trial with select customers is expected to run for six to twelve weeks.

Sources

• New York City Artificial Intelligence
• New York City Artificial Intelligence
• AI whistleblowers encourage NYC Council to enforce more regulation
• City Council takes on debate over catastrophic risks of AI
• NYC Lawmakers Aim to Lead Nation in Establishing AI Safety Rules
• Mistral CEO says new AI model beats Chinese ones in some areas
• Mistral AI unveils new AI model aimed at 'narrowing the gap' with top Chinese competitors
• Advancing computer use with Ironclad
• Advancing computer use with Ironclad
• Can Agent Harnesses and Inference Engines Hear Each Other? The HEAR Protocol for Agentic LLM Serving
• OpenAI Reportedly In Talks With UAE Funds For $30B Funding – Even As Its AI Agents Raise Fresh Concerns
• Zscaler Climbs 3% as Investor Day Reaffirms Guidance in Agentic AI Pitch; CrowdStrike and Palo Alto Tick Up
• Thomas College gets $3 million federal AI grant
• Maydin Media Expands Award-Winning KAGAPSO® Business Growth Framework with AI-Powered Revenue Systems
• Beyond Instruction Following: Learning Grounded Skill-Following with Skill Contracts
• Orgvue accelerates AI-powered workforce design with innovative WebMCP interface