When AI Becomes the Operating System
Artificial intelligence is evolving from an assistance system to an increasingly autonomous actor in digital systems. The Agentur für Innovation in der Cybersicherheit GmbH (Cyberagentur) is taking a fundamental step further with “Artificial Intelligence as Operating System” (AI-as-OS). At a digital partnering event on September 2, 2026, from 9 a.m. to 12 p.m., it will bring together experts to network expertise for research into AI-native operating systems and to initiate potential research partnerships for the planned research program.
The technological transition from generative to agent-based artificial intelligence is transforming the relationship between humans, software, and computer systems. AI systems no longer merely generate content; they can plan tasks, deploy tools, and increasingly carry out actions within digital systems autonomously. At the same time, AI functions are being integrated more deeply into end devices and system software.
With “AI-as-OS,” the Cyberagentur is addressing this development and taking it a step further. Artificial intelligence should not be viewed merely as an application or assistance system within existing operating system structures. Rather, the aim is to investigate whether AI itself can become the central control element of a computer system.
Such an AI-native operating system could manage system resources—such as processor power and RAM—in a data-driven, adaptive, and context-dependent manner. It could control processes and react dynamically to changes in the system’s state. AI could also be directly integrated into mechanisms for detecting and defending against cyberattacks.
This also fundamentally changes the security landscape. The more AI modules decide at levels close to the system core, the more important traceability, controllability, and robustness become. At the same time, the question arises: what decision-making powers can be delegated to such a system, and how can human control be technically ensured?
“With AI-as-OS, we are deliberately posing a question for which there is still no ready-made technical answer today: What happens when artificial intelligence is no longer a guest in the operating system but instead takes over its fundamental control functions itself? It is precisely at this boundary between theoretical possibility and technological feasibility that disruptive research begins for us,” explains Dr. Philippe Krajsic, Research Officer for Key Technologies at the Cyberagentur.
The research program focuses on three interconnected areas: the development of AI-centric operating system architectures; the integration of AI-native methods into core functions such as resource management, process control, and memory organization; and new approaches to control, security, and stability.
This research question is gaining additional significance due to the rapid development of autonomously acting AI systems. Internationally, there is a growing focus on the security, identity, authorization, and controllability of AI agents. The challenge is thus shifting from the question of what content an AI model can generate to the question of what actions an AI system is permitted to perform independently and how these can be technically controlled.
AI-as-OS operates one level below this. Today’s AI agents rely on existing operating systems, interfaces, and authorization models. In contrast, this research program investigates architectures in which AI itself becomes an integral part of the system level that controls resources, processes, and security mechanisms.
Whether such an architecture is technically feasible, securely controllable, and superior to existing operating system concepts remains an open question. It is precisely this uncertainty that forms a significant part of the research focus. The aim is to gain scientific insights into the architectural, security, and control principles required by AI-native operating systems and to identify their technological limitations.
Partnering Event Brings Together Diverse Expertise
Research into AI-native operating systems requires collaboration across various scientific and technological disciplines. With the digital partnering event on September 2, 2026, from 9 a.m. to 12 p.m., the Cyberagentur is creating a platform for this at an early stage.
The event is aimed in particular at stakeholders from AI research, operating system and system software development, IT security, computer architecture, resource and memory management, and related research fields. Participants will receive further information on the research questions and aims of AI-as-OS. At the same time, they can present their own areas of expertise and meet potential partners for forming interdisciplinary research consortia.
The Partnering Event will take place digitally on September 2, 2026, from 10 a.m. to 12 p.m.