ModelOp Unveils Industry-First Full-Stack AI Policy Enforcement With NVIDIA OpenShell
New NVIDIA OpenShell integration enforces AI policy at the operating system, making ModelOp the first AI governance
Press Release Disclaimer: This is a press release distributed through the XPR Media network. It has not been independently verified by our newsroom.
PARK CITY, Utah, Nov. 04, 2026 (GLOBE NEWSWIRE) — A critical governance gap is emerging as enterprises race to deploy autonomous AI: an agent can be fully approved, compliant on paper and still take actions that violate corporate policy. Today, ModelOp, the Enterprise AI Command Center for Fortune 500 and Global 2000 enterprises, announced its Full-Stack AI Policy Enforcement to close this gap. ModelOp’s Full-Stack AI Policy Enforcement centralizes and continuously manages AI policy in ModelOp’s system of record and enforces it at four levels: process, execution, operating system (OS) and network.
ModelOp’s new integration with NVIDIA OpenShell, NVIDIA’s open-source secure runtime for AI agents, extends existing AI governance enforcement to the OS, where policy governs what an agent can read, write, run and send.
“The biggest misconception in enterprise AI governance is that approving an agent means it will adhere to all of your policies in production,” said Dave Trier, CEO of ModelOp. “An agent can pass every risk assessment and approval and still violate policy the moment it starts operating. If your governance ends at deployment, you’re not actually enforcing your policy. You’re documenting what you hoped it would do. ModelOp bridges this gap. We already enforce policy across workflows, execution platforms and gateways. Now with NVIDIA OpenShell, we’re extending it to the operating system, so enterprises can enforce policy wherever their agents act.”
Why AI Governance Must Be Enforced at the Operating System Level
Unlike traditional software, AI agents act on their own. They execute code, invoke tools, write and move data, access sensitive files and communicate with external systems. A policy that ends at approval depends on the agent behaving as expected.
In reality, policy has to hold at four levels: the process that governs an agent’s end-to-end lifecycle, the platform that executes it, the operating system it runs on, and the network it connects to. The process level alone spans intake, risk tiering, risk-proportional controls, approvals, evidence, and release gating, and more. In many enterprises, each level is managed by a different team in a different tool (or, in the worst case, not managed at all), often with no shared record or transparency into what policies are enforced across the disparate layers, teams, and agent platforms.
When management and enforcement aren’t consistent and connected across those levels, an agent can slip through the gaps between them. An AI gateway can block an unapproved API call, for example, but an agent that can run code can write its own path to the same destination unless the operating system stops it.
Four Levels of AI Policy Enforcement
ModelOp closes the gaps between levels by centrally and continuously managing policy and connecting enforcement across all four. It already enforces policy at the process, execution and network levels through its lifecycle workflows and integrations with execution platforms such as Microsoft Agent 365 and Amazon Bedrock AgentCore and AI gateways such as Kong. With NVIDIA OpenShell, it now enforces policy at the operating system as well, using deny-by-default, kernel-level controls that run outside the agent’s own process.
Bringing Enterprise-Wide Governance to NVIDIA OpenShell
ModelOp turns OpenShell sandbox permissions from isolated technical settings into governed policy, tied to each agent’s business purpose, owner, risk tier and approved version and carried from one agent version to the next. Powered by MADE™, the ModelOp AI Delivery Engine, the integration works in three steps:
- Manage: Security, IT and AI governance teams set global OpenShell policies in ModelOp, solution owners add the policies their use case needs, and ModelOp records the policy each agent version runs under.
- Enforce: ModelOp registers OpenShell gateways, tracks which agents run in which sandboxes, keeps approved policy in sync and shows each agent version’s compliance status.
- Observe: When OpenShell blocks an action, ModelOp routes the exception and suggested policy change through configurable review workflows, applies approved changes to the running sandbox and records every decision for audit.
“Each enforcement point sees one slice of what an agent does,” said Jim Olsen, CTO of ModelOp.”The gateway sees API calls, the sandbox sees files, processes and connections, and the execution platform sees version and identity. ModelOp keeps the policy for all of them in one system of record, version by version, so a blocked action reaches the right reviewers with the agent’s owner, use case and risk tier attached.”
For CIOs, CTOs and CISOs scaling agents, ModelOp’s Full-Stack AI Policy Enforcement provides central, continuous and consistent AI policy management, enforced at every level of the stack, including the operating system.
To learn more, check out ModelOp CEO Dave Trier’s post on the ModelOp blog here.
About ModelOp
ModelOp is an AI Delivery Platform for enterprise AI leaders: an Enterprise AI Command Center and system of record that powers workflows, enforces governance, and generates operational intelligence in one operating layer to industrialize ML, GenAI, Agentic AI and vendor AI above AI development and execution, security, GRC, ITSM and data management. It’s used by the most complex and regulated institutions in the world, including major banks, insurers, regulatory bodies, healthcare organizations, government, defense, manufacturing, energy, global CPG and retail companies.
ModelOp connects and extends existing investments in GRC, ITSM, security and data management without replacing them, so enterprises can accelerate AI from idea to production and deliver AI rapidly, safely and profitably.
ModelOp is a member of the NVIDIA Inception Program.
Follow ModelOp on LinkedIn or visit modelop.com to learn more or schedule a demo.
Media Contact
Ria Romano, Partner
RPR Public Relations, Inc.
786-290-6413
A photo accompanying this announcement is available at https://www.globenewswire.com/NewsRoom/AttachmentNg/8710280e-6e2d-41c8-acdd-813a3e38748a
