Here is today's AI news. Story 1. OpenAI. OpenAI reportedly shelved a new AI model over safety concerns after it showed poor ability to follow instructions.. OpenAI has reportedly shelved GPT-6.1 Astra before release after internal testing found it fell short on safety and instruction-following. Safety-systems head Saachi Jain said it “didn’t quite meet the bar” for staying within scope and authorization, or explaining what work it had done. The decision underscores a growing industry challenge: as AI agents take on more complex tasks, capability gains must be matched by reliable safeguards and transparency. Story 2. AMD. AMD will acquire Fei-Fei Li’s World Labs for $8.2 billion, with Li joining AMD as executive vice president and chief scientist.. AMD has agreed to buy World Labs in an all-stock deal valued at about $8.2 billion, bringing AI researcher Fei-Fei Li to the chipmaker as executive vice president and chief scientist. The deal, expected to close by year-end pending approvals, gives AMD a foothold in “spatial intelligence”: models that generate and simulate interactive 3D worlds from text, images and video. The research could support robotics and other physical-AI systems, while helping AMD shape its hardware and software for future AI workloads. Story 3. Anthropic. Anthropic released Sonnet 5.5, describing it as a faster, lower-cost model for work tasks.. Anthropic released Claude Sonnet 5.5 on Sept. 28, positioning it as a faster, more efficient model for coding and everyday knowledge work, including creating documents, slides and spreadsheets. The company says it generates outputs more than 30% faster and can cost up to 30% less per task than Sonnet 5, largely by using fewer tokens. API rates remain $2 per million input tokens and $10 per million output tokens. The update could make AI agents cheaper to run at scale. Story 4. Nvidia. Nvidia launched a platform designed to help companies control and monitor AI agents.. Nvidia launched its Open Agent Safety Platform on Sept. 28 to help companies govern AI agents from testing through deployment. It pairs open-source OpenShell, which sets runtime limits on agents’ access and actions, with Sentry, a watchdog designed to run on BlueField-4 data-processing units and monitor agents independently. Nvidia says Sentry can quarantine agents that breach their boundaries in milliseconds, adding a hardware-based security layer as businesses adopt autonomous AI. Story 5. Meta. Meta launched an enterprise AI platform and appointed MongoDB CEO Dev Ittycheria to lead the initiative.. Meta launched the Meta Enterprise Platform on Sept. 28, bringing its AI tools to businesses and developers. The initial lineup includes the Muse agent, Meta Business Agent, Muse API and Muse Code. MongoDB CEO CJ Desai—not Dev Ittycheria—will lead the effort as Meta’s chief enterprise platform officer, reporting to Mark Zuckerberg. The move puts Meta in the race to sell companies deployable AI models, agents and coding tools; MongoDB named Ittycheria interim CEO. Source: [Meta](https://about.fb.com/news/2026/09/launching-meta-enterprise-platform/), [Reuters](https://www.reuters.com/technology/mongodb-ceo-desai-steps-down-lead-metas-enterprise-platform-2026-09-28/). That's all for today's AI news briefing.