Security Warning for Open-Weight Artificial Intelligence Models from Microsoft and Nvidia

Microsoft and Nvidia published a joint paper aimed at evaluating security vulnerabilities during the deployment of open-weight artificial intelligence models. The companies warned businesses that changes in model weights and configurations could weaken security measures.

In a joint report published last week, 30 types of risk across five categories were identified, along with a process for their evaluation, mitigation, and monitoring. The guide stipulates that businesses verify model provenance, test the configurations they intend to deploy, and reassess risks following significant changes. The report emphasized that weight openness alone does not determine the risk of an artificial intelligence system, as risk depends on the model, use case, deployment, and safeguards.