AI News from September 16, 2026

NVIDIA says TensorRT Edge-LLM ran Qwen3.6-27B 6.4x faster on Jetson AGX Thor, highlighting cache and quantization gains for edge agents.

Anthropic and OpenAI propose embedding independent AI safety evaluators, but researchers say access, disclosure rights, and regulation will determine oversight.

Google is opening a Google Home MCP server to AI agents, enabling natural-language device control and camera-history access while limiting early access to U.S. subscribers.

Hang Ten Systems, founded by former Infosys CEO Vishal Sikka, raises $53 million to expand AI-led enterprise software delivery and consulting.

OpenAI is testing Sponsored Agents in ChatGPT and expanding ad tools with HubSpot and Shopify, bringing conversational commerce closer to deployment.

OpenAI confirms weeks of AI safety talks with Anthropic and Google DeepMind, sharpening a debate over coordination, antitrust, and U.S. policy.

Morgan Stanley reportedly puts China’s consumer AI opportunity near 300 billion yuan, with Tencent, Alibaba and Meituan positioned to benefit.

Meta’s new WhatsApp Business MCP server lets AI coding agents automate setup, templates, testing, and troubleshooting for business messaging teams.

Nvidia CEO Jensen Huang says AI safety should be engineered by companies and existing law, rejecting new regulation and renewing debate over accountability.

Profound raised $180 million at a $1.8 billion valuation, accelerating the race to help brands rank in AI-generated search results as enterprise demand grows.

Anthropic’s Fable retention policy has triggered enterprise pullbacks, exposing why zero-data-retention promises have not resolved AI trust risks.

Google DeepMind launches Gemini 3.8 Live audio models with low API pricing, challenging OpenAI’s GPT-Live-1 on cost, quality and developer adoption.