[AI-as-OS] Artificial Intelligence as Operating System

Artificial intelligence (AI) methods have been expanding into new areas of application for years. AI systems are increasingly capable of handling complex tasks with greater autonomy—for example, through agent-based approaches that run an AI agent on a computer to perform any task automatically or even autonomously, sometimes without explicit human instructions but rather based on identified intentions (see, for example, the open-source software OpenClaw, which can handle tasks such as calendar maintenance or file management). Just how deeply AI algorithms and models are now integrated into everyday applications is also evident in the fact that end devices are increasingly operable entirely without visible applications and exclusively via voice control, which fundamentally rethinks the way we interact with them (such as Deutsche Telekom’s AI Phone). The central control units of such end devices—the operating systems themselves—are also increasingly being enhanced with new AI functionalities to continuously improve system efficiency. These include initial theoretical—and in some cases already practical—approaches to using the latest AI technologies for file system or storage management. These developments mark a fundamental paradigm shift: AI is no longer viewed merely as an add-on, but increasingly as a native architectural component of the operating system itself. The AI-as-OS program aims to explore how artificial intelligence can be understood not merely as a separately encapsulated functionality that is simply executed by the operating system, but as the operating system itself—Artificial Intelligence as Operating System. Behind this lies the vision of an AI that not only supports individual applications but also acts as the central control system for the entire computer ecosystem, managing system resources (including the CPU, RAM, and system security) in a data-driven, adaptive, and context-aware manner. The aim of the program is to make fundamental contributions to the development of AI-based operating systems as the control unit of a computer system. Through the extensive integration of AI technologies into deep operating system layers, such a system is intended to manage system resources, enable context-aware operational decisions, and enhance system security through adaptive threat detection and defense. Since the underlying technology is still largely in a speculative stage, the scope of development work ranges from the general architecture of such an operating system to the integration of AI-based core mechanisms, and includes new considerations regarding the security and stability requirements for such an operating system, depending on the target platform (e.g., desktop, data center, edge device) and the application environment. The program’s main areas of research include: • AI-Centric Operating System Architectures: Concepts for the fundamental design of an operating system in which AI is viewed not as an add-on feature but as an integral part of the architecture. • AI-Native Core Mechanisms: Integration of AI-based methods into core operating system functions such as resource management, process control, and memory organization. • Control, Security, and Stability: Requirements for traceability, controllability, and robustness of an AI-based operating system, as well as approaches to adaptive threat detection and defense. From today’s perspective, it remains to be seen whether—and to what extent—the vision of an AI-native operating system can be technically realized. This is precisely where the program’s value lies: AI-as-OS explores a largely speculative field at the intersection of AI research and system software, the results of which could set the course for the operating systems of tomorrow.