AI news Nov 18: enterprise agents, major funding and tougher oversight
Microsoft has announced a broad set of agent-focused products and Azure services, while California is proposing stronger controls for artificial intelligence. Isomorphic Labs has also announced a $5 billion Series C funding round, adding to a busy day for enterprise AI and infrastructure.
- Microsoft expands Copilot, Foundry and Azure for enterprise agents.
- Isomorphic Labs announces a $5 billion Series C round.
- California proposes shutdown powers and independent AI oversight.
- OpenAI reports gains in chip and processor design.
- Local AI hardware and inference capacity continue to expand.
Microsoft puts agents at the centre of Ignite
Microsoft outlined a broad set of AI and cloud announcements at Ignite, focused on collaboration between people and software agents, task-specific models and enterprise control. The plans include Copilot Workspaces, new models in Foundry, Agent 365 Business Analytics and expanded Azure infrastructure.
Microsoft also expanded Foundry Agent Service with generally available real-time voice and browser automation, alongside enterprise deployment controls and Microsoft 365 integration. The Foundry Browser Automation tool is now generally available, while Foundry Toolboxes let developers reuse and version tools, skills and agents across production systems.
Other Microsoft announcements include new networking, governance and deployment controls for hosted agents, a broader frontier AI ecosystem built around Copilot, agents and Azure, and an expanded partnership with Intel for Azure infrastructure and AI demand. Factory is also expanding its Microsoft partnership to deploy self-improving software factories in Azure.
Isomorphic Labs announces $5 billion financing
Isomorphic Labs has announced a $5 billion Series C funding round, according to a PR Newswire release dated November 18, 2026. A second release identifies the financing as a $5 billion Series C, but the supplied announcements provide no further transaction details.
In related enterprise infrastructure news, AWS has added managed NVIDIA Dynamo inference to SageMaker HyperPod. The integration is designed to serve large language models across multiple nodes and separate prompt processing from token generation.
California proposes stronger AI controls
California Governor Gavin Newsom is proposing mandatory AI safety standards, independent oversight and emergency shutdown authority. The recommendations would also place limits on recursive self-improvement and give the state a larger role in testing and monitoring artificial intelligence.
Research raised separate questions about AI personalisation. A preliminary Princeton field audit of 160 US users found that ChatGPT’s memory settings can affect the partisan slant of election-related answers, although direct voting recommendations were rare.
Chips, local models and applied research
OpenAI says its models produced silicon-proven improvements to its Jalapeño inference chip and up to 14% gains against optimized baselines on AMD hardware. The company also reported stronger results on open-source RISC-V processor design tasks.
NVIDIA RTX Spark compact desktops from ASUS, Dell and MSI are now available, with local agents, image and video generation, and model fine-tuning. Dell’s Creator Edition Desktop can run models of up to 200B parameters on-device. Dell also announced private cloud AI options with Microsoft, AMD Instinct MI350P configurations and a liquid-cooled networking rack in production at Nscale.
Other developments include Google DeepMind’s Genie 4 rollout in Project Genie, Workday’s planned acquisition of Honeydew, and Dropbox’s completed acquisition of Tigris Data. Epoch AI reported a 4.6× increase in estimated agentic inference capacity between Q2 2025 and Q2 2026.


