Position: It is Time to Virtualize Foundation Models with a Self-evolving Operating System Layer

arXiv:2609.19203v1 Announce Type: new Abstract: AI applications have shifted from single, monolithic foundation models (FM) to compound agentic systems. Yet today's stacks remain fragmented: even as protocols (e.g., MCP, A2A) ease tool/agent connectivity, each framework embeds an implicit runtime for state, memory, budgets, and guardrails, making behavior non-portable and governance brittle. It mirrors computing before operating systems, when every program re-implemented basic services. This position paper argues that the field now needs a Foundation Model Operating System (FMOS) -- a system layer that virtualizes FM interactions analogous to how virtual machines abstract physical hardware, giving applications the illusion of dedicated, trustworthy FM instances with effectively unbounded capabilities. Internally, the FMOS orchestrates knowledge across memory tiers, model selection and resource allocation, and verification and policy enforcement. Like the human brain switching between fast intuition and slow deliberation, the FMOS learns when to intervene and when to let inference proceed directly and continuously adapting its policies based on operational experience.

Proposes a new system layer for AI applications

The proposal for a Foundation Model Operating System could improve AI system interoperability, governance, and resource use, enabling more scalable and trustworthy AI applications. However, it also consolidates control and may introduce new complexity and vulnerabilities. The implications could be mixed depending on implementation and governance.

This is a conceptual architecture proposal, not an empirical result, with no deployed system or measured outcomes. Implications for governance and safety could go either way depending on implementation, so the societal effect is genuinely uncertain and thinly evidenced.

This paper proposes a new architectural layer for AI systems, the Foundation Model Operating System (FMOS). Its claims are primarily technical about improving efficiency, portability, and governance of compound AI agentic systems. The proposed system does not inherently present utopian or dystopian societal implications.

The paper proposes a new operating system layer for AI that could greatly improve efficiency and governance. The FMOS could make AI systems more adaptable and trustworthy, leading to significant benefits for humanity.

SOURCE     arXiv cs.AI (tier 1)
   ↓
DOCUMENT   58d394a4-0d12-4cdc-844b-8835022b2780
           https://arxiv.org/abs/2609.19203
   ↓
EVIDENCE   2 extracted excerpts
   ↓
MODEL RUN  5 runs, methodology 1.0.1
   ↓
SCORE      +3  (Mixed or uncertain)
   ↓
CONFIDENCE 64%