Microsoft Copilot Gains Expanded Control Over Windows Files and System Actions via Hybrid Intelligence

Microsoft Copilot Gains Expanded Control Through Hybrid Intelligence

Microsoft unveiled significant upgrades to its Copilot AI system at the Windows and Surface event, introducing a framework called Hybrid Intelligence that grants Microsoft Copilot unprecedented control over Windows files and system operations1

1

. The new capabilities allow Copilot to access and process local files while executing actions across the operating system, fundamentally changing how users interact with Windows AI features2

2

.

Built atop Microsoft's HydraFusion model router, Hybrid Intelligence enables Windows apps and Copilot features to intelligently switch between local AI models for simple or privacy-sensitive tasks and cloud AI models like OpenAI's GPT-6 or Anthropic's Claude for complex requests2

2

. Jacob Andreou, Microsoft's EVP of Copilot, demonstrated the system's potential through a tax filing scenario where Autopilot searched across different folders, renamed files, created a zip archive, and drafted an email with attachments—all from a single natural language request1

1

.

Enhanced Security Through Execution Containers

To address privacy concerns that have plagued Microsoft since the first Copilot+ PCs arrived in 2024 alongside Qualcomm's Snapdragon X Elite SoCs, the company introduced Microsoft Execution Containers (MXCs)2

2

. These agent sandboxes apply fine-grained, user-controlled permissions at runtime, with Pavan Davuluri, Microsoft's EVP of Windows and Devices, claiming Windows is now "the most secure platform for agents"2

2

.

Microsoft emphasized that agentic capabilities in Copilot will only access files when explicitly requested by users. The company demonstrated business-focused protections, showing how AI models can detect and report prompt injection attacks on destination websites, with businesses able to configure different permissions for both users and their agents3

3

. On-device processing for sensitive tasks aims to reduce load on Microsoft's servers while keeping low-priority requests local2

2

.

AI-Powered Taskbar Transforms Windows Search Experience

Microsoft showcased a new Windows search experience integrated into the AI-powered taskbar, addressing long-standing limitations in Copilot's ability to perform system-level tasks4

4

. The enhanced search allows users to execute quick actions directly from the taskbar, including toggling dark mode, adjusting microphone volume with inline sliders, sending text messages, enabling Do Not Disturb, taking screenshots, and arranging desktop layouts1

1

4

4

.

The new Windows search experience will be available starting Fall 2024 on Windows 11 PCs, according to Windows and Surface boss Pavan Davuluri1

1

. Users can ask questions through the taskbar search and open responses in the main Copilot app for more detailed context when needed4

4

.

Expanded Framework Support and Third-Party Integration

To support Hybrid Intelligence functionality, Microsoft extended Windows ML—its inference framework built on the ONNX runtime—to include Llama.cpp, one of the highest-performing inference engines for local AI inference2

2

. This integration should enable Llama.cpp to leverage a broader range of accelerators for improved performance.

Microsoft is rolling out Hybrid Intelligence to GitHub Copilot next week and to the Copilot app over the next couple months, Andreou confirmed1

1

2

2

. The company also introduced a new onboarding experience for deploying third-party agent harnesses like OpenClaw and Hermes, featuring a streamlined setup flow that lets users select models, install them, and begin working immediately2

2

4

4

.

Code Generation and Autopilot Workflows

Microsoft unveiled a new Code section within Copilot that enables users to generate Windows-native apps with single text prompts, requiring no developer experience3

3

4

4

. The feature runs in secure MXC local sandboxes and can delegate work to other local models when appropriate.

The Autopilot feature showcased the full potential of Hybrid Intelligence, demonstrating automated workflows where the AI searches the PC, locates requested files, attaches them to emails, drafts messages, and waits for user permission before sending3

3

4

4

. Microsoft also demonstrated third-party integration with Perplexity, showing an agent resizing photos, renaming them, and creating a local HTML gallery with light and dark mode support—all processed locally without cloud uploads4

4

.

The announcement came alongside the reveal of an Nvidia N1X-powered Surface Laptop, positioning Microsoft to compete more aggressively in the AI-enhanced computing space2

2

. However, concerns remain about the impact of running these models locally on battery life, fan noise, and overall workload quality for many users2

2

.