Microsoft and NVIDIA publish joint framework (Risk Management for Open-Weight Models in Enterprise) for enterprise open-weight model risk

On Oct 16, Microsoft Security Blog flagged a new joint Microsoft-NVIDIA whitepaper and companion repository ('Risk Management for Open-Weight Models in Enterprise') giving enterprises a shared taxonomy, evaluations and controls for open-weight model risk, including guidance on evaluating open-weight vs closed models and securing deployment.

Open-weight models are spreading into enterprise and sovereign environments faster than governance; this is a rare joint taxonomy/controls reference from two of the biggest AI-infra vendors — useful as an evaluation baseline for AI-SPM and model-governance decisions, and as a signal of demand for model-scanning/vetting tooling.

Enterprise AI/platform and security teams evaluating open-weight deployments (incl. sovereign/air-gapped) should adopt the taxonomy and controls as a baseline for procurement and AI-SPM policies.