Industry Leaders Unite in Open Secure AI Alliance for AI Safety and Security

On July 27, 2026, NVIDIA, together with the Linux Foundation, OpenSSF, and dozens of industry leaders, announced the formation of the Open Secure AI Alliance. The alliance aims to build and share open-source tools to promote responsible use and trust in AI. It believes that open-source software is a key pillar of the global economy, and cybersecurity is one of its greatest beneficiaries.

The alliance's mission is to ensure defenders have trusted and controllable open frontier tools. Its core argument is that in AI security, open models and tools are essential for democratizing defense capabilities, increasing transparency, protecting data, and avoiding single points of failure.

The alliance's formation is partly driven by lessons from the recent Hugging Face security incident. In that incident, closed AI tools could not distinguish between attackers and defenders, hindering necessary forensic analysis. Ultimately, Hugging Face ran the open-source GLM 5.2 model on its own infrastructure, analyzing over 17,000 action records and successfully containing the intrusion.

This incident shows that when defenders cannot inspect, adapt, and run advanced AI on their own infrastructure, their response capabilities are limited at the moment when speed is most needed. Therefore, companies and countries need open frontier defense tools and technologies so that critical industries can build secure systems across multi-vendor ecosystems.

The alliance acknowledges that open models can be misused (e.g., to weaken security measures or for cyberattacks), but argues that these risks are not unique to open systems and exist in closed systems as well. The correct response is not to deny defenders access to open systems, but to combine openness with strong security measures, clear abuse rules, rigorous evaluation, and rapid fixes.

The alliance advocates that the world needs both closed and open models to work together, so defenders can choose the right system for the task and ensure transparency, adaptability, and sovereign control when security demands it.

NVIDIA contributed the new open-source NVIDIA Labs Object-Oriented Agent (NOOA) research framework, designed to make AI agent behavior easier to test, trace, audit, and govern. The framework is now available on GitHub.

Other member contributions include: HPE's SPIFFE/SPIRE zero-trust identity framework for cryptographically verifying AI agents and services; Hugging Face's SafeTensors secure tensor format providing transparency and no remote code execution guarantees; IBM and Red Hat's Lightwell extending open-source supply chain security with digitally signed patches; Microsoft's MDASH multi-model agent scanning tool coordinating specialized AI agents to discover and prove exploitable vulnerabilities; SpaceXAI open-sourcing the Grok Build terminal AI coding agent and planning to open-source Grok model weights.

The alliance calls on policymakers and regulators to view open models, tools, and security tools as defense assets, not liabilities. Blanket restrictions on open frontier AI systems would weaken defense capabilities and could lead to power, dependency, and vulnerability concentrated in a few closed providers.

The alliance recommends that companies and governments invest in shared open AI defense infrastructure, including datasets, evaluation frameworks, attack simulators, and red-teaming tools, similar to past investments in open-source software.

The alliance believes that the era of AI agents can be an era of resilience and shared security. With the right choices, open secure AI systems can provide defenders with the tools they need, enhance competition, extend technological leadership, and ensure AI security is built in an open environment.

The alliance invites governments, industry, and researchers to join in building AI systems that can withstand scrutiny, be flexibly improved, and are open enough to mobilize the entire defender community.