Microsoft is moving more artificial-intelligence work from the cloud onto personal computers, unveiling a coding model that can run locally and new Windows security controls designed to limit what desktop AI agents can access.
The company announced the products Wednesday at an event in San Francisco, according to Reuters. Microsoft also showed a Surface Laptop Ultra powered by Nvidia’s new RTX Spark chips, positioning Windows PCs as machines that can run more capable AI workloads without sending every task to remote data centers.
What Microsoft announced
The local coding model is intended to perform software-development tasks directly on a compatible PC. Local processing can reduce latency, keep some information on the device and lower reliance on metered cloud computing, although actual performance will depend heavily on hardware, memory and software support.
Microsoft also introduced Microsoft Execution Containers, a mechanism intended to isolate AI agents and prevent unauthorized access to data or actions on a laptop or desktop. The company said Anthropic, OpenAI and Nvidia plan to use the technology. Microsoft CEO Satya Nadella described the goal as making the desktop a secure environment for agents to execute.
Execution controls are increasingly important because AI agents can do more than generate text: they may open files, run code and interact with other applications. Microsoft previously described its Microsoft Execution Container SDK as an operating-system-level control layer using process and session isolation. Wednesday’s announcement expands that security story as local AI use grows.
New hardware—and a cost barrier
The Surface Laptop Ultra uses Nvidia silicon designed for demanding local AI work. Microsoft and Nvidia are trying to broaden the market for high-performance Windows AI PCs, but the approach has a practical constraint: memory and advanced chips remain expensive.
That means the most capable on-device models may initially be limited to developers, companies and buyers willing to pay for premium hardware. Microsoft’s claims about performance, privacy and security will also need independent testing once the products and software are broadly available.
What it means for PC users
The announcement points toward a hybrid model in which some AI tasks stay on-device while larger or more demanding jobs use cloud services. For users, the key questions will be which computers qualify, what data agents can reach by default, how clearly Windows displays permission requests and whether isolation controls prevent real-world attacks.
AskNovus will track release details and independent security testing in Artificial Intelligence and Consumer Technology.
Featured image: FlyD via Unsplash.
Sources
- Reuters — Microsoft brings more AI to PCs (October 7, 2026)
- Microsoft Security Blog — Execution Container and agent-security background (June 2, 2026)