NVIDIA and Partners Launch Open Secure AI Alliance

Technology companies and open source organisations have formed the Open Secure AI Alliance to develop and share tools for securing artificial intelligence systems and agents. NVIDIA announced the group on 27 July 2026 and named inaugural partners from cloud computing, cybersecurity, enterprise software, open source foundations, and AI research. The list includes Microsoft, IBM, Red Hat, Hugging Face, and the Linux Foundation.

The alliance says contributors are building an open defence stack covering identity, permissions, isolation, model storage, vulnerability scanning, secure coding, logging, and evaluation. Its announcement points to existing projects including SPIFFE/SPIRE, Safetensors, Lightwell, and MDASH. It presents these as member contributions and does not identify a jointly released alliance product or delivery schedule.

NVIDIA also released the open source NVIDIA Labs Object-Oriented Agent framework, known as NOOA. The framework represents an agent as a Python class whose fields hold state, methods expose capabilities, docstrings provide prompts, and type annotations define contracts. Its documentation says models can act by writing Python and warns that in-process validation is not a containment boundary.

The documentation instructs users to run agents that execute generated code inside operating-system isolation such as a container, virtual machine, or NVIDIA OpenShell. The Drupal connection remains indirect: the Drupal AI project supports provider integrations and agents, while its observability tooling logs calls to AI providers. No Drupal project or organisation appears in NVIDIA’s partner list, but the NOOA warning gives Drupal teams a practical baseline: logging and agent testing do not replace restricted permissions or isolated execution where an agent can execute generated code or alter site state.