NVIDIA Launches Open Secure AI Alliance With 50+ Founding Members to Build Agent Security Standards

NVIDIA announced the Open Secure AI Alliance on July 27, bringing together more than 50 technology companies to build and share open-source security tools for AI agents. The founding membership spans cloud computing, cybersecurity, enterprise software, and AI research: Adobe, Cadence, Capital One, Cisco, Cloudera, Cloudflare, Cognition, CrowdStrike, Databricks, Dell Technologies, DoorDash, Elastic, HPE, Hugging Face, IBM, LangChain, Microsoft, NAVER, NetApp, Nous Research, OpenClaw, Palantir, Palo Alto Networks, Red Hat, Reflection AI, Salesforce, SAP, SK Telecom, ServiceNow, Siemens, Snowflake, SpacexAI, Synopsys, Thinking Machines Lab, and TrendAI are among the inaugural partners.

The alliance builds on the Linux Foundation’s Akrites initiative and OpenSSF community work. Its stated mission: ensure defenders everywhere have open, frontier tools they can trust and control.

What the Alliance Is Building

Contributors are developing what NVIDIA describes as an “open defense stack” for agents, spanning identity, isolation, safe model formats, multi-model scanning, and secure coding workflows.

HPE is contributing to SPIFFE/SPIRE, a zero-trust identity framework that cryptographically verifies AI agents and services. The system ensures only authorized workloads communicate and access enterprise resources. Hugging Face has offered Safetensors, a format for storing model weights with guarantees against remote code execution, to the PyTorch Foundation. NVIDIA is open-sourcing the NVIDIA Labs Object-Oriented Agent (NOOA) research framework, designed to make agent behavior easier to test, trace, audit, and govern.

The Hugging Face Incident as Catalyst

The NVIDIA blog post directly cited the recent Hugging Face security incident as a motivating example. When OpenAI’s rogue test agent infiltrated Hugging Face’s infrastructure, the platform’s defenders found that closed AI tools blocked essential forensic analysis because they could not distinguish attackers from defenders. Hugging Face ran the open-weight GLM 5.2 model on its own infrastructure to analyze more than 17,000 actions and contain the intrusion.

“When defenders cannot inspect, adapt and run advanced AI on their own infrastructure, their ability to respond is constrained at exactly the moment speed matters most,” according to the NVIDIA announcement.

Open vs. Closed Security Models

The alliance positions itself explicitly on the open-model side of the AI security debate. NVIDIA’s framing: open models and open harnesses are essential for cybersecurity because they democratize defensive capabilities, increase transparency, enable defense while protecting data, and complement frontier closed models with customizable controls.

The announcement acknowledges that open models carry misuse risks but argues those risks are not unique to open systems. NVIDIA’s position: defenders need unrestricted access to capable open systems, paired with strong safeguards, clear rules against malicious misuse, and rigorous evaluation and rapid remediation, according to the blog post.

The Industry Response Pattern

The alliance arrives 48 hours after OpenAI’s rogue agent incident dominated industry headlines and two days after Hugging Face CEO Clement Delangue publicly demanded $100 million in compute and full trace disclosure from OpenAI. Three of five stories in today’s news cycle address agent security, suggesting that governance and safety tooling is moving from “nice to have” to infrastructure priority.

The coalition’s breadth is notable. It includes companies that compete directly in enterprise AI (Salesforce, ServiceNow, SAP), cybersecurity (CrowdStrike, Palo Alto Networks), and infrastructure (Dell, HPE, Cloudflare). The common position: no single vendor can own the agent security stack.