Nvidia and Microsoft have launched the Open Secure AI Alliance, a coalition aimed at developing and sharing open-source security tools for artificial intelligence systems.
The alliance brings together technology giants, including SpaceX, IBM, Adobe, CrowdStrike, Hugging Face, Cisco, Dell Technologies, Cloudflare, Salesforce, Siemens and Palo Alto Networks.
The group was announced on Monday, July 27, 2026, days after an autonomous AI agent linked to OpenAI was reported to have carried out an intrusion at AI startup Hugging Face.
Nvidia said the alliance will focus on helping developers and security teams inspect, adapt and run AI models on their own infrastructure.
The company also said it will contribute open model weights, training data and research on AI agent systems to the project.
A part of Nvidia’s contribution is the release of its Object-Oriented Agent framework through an open-source project hosted on GitHub.
The formation of the alliance follows a security incident involving Hugging Face, where the company said an autonomous AI agent carried out an end-to-end intrusion. According to Nvidia, the incident exposed challenges around investigating attacks involving advanced AI systems.
Nvidia said Hugging Face turned to an open-weight Chinese AI model, GLM 5.2, to review more than 17,000 actions linked to the incident after some closed AI models could not support its forensic investigation.
“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,” Nvidia said in a statement.
Currently, technology companies are divided over whether these systems should face more restrictions.
Nvidia and other supporters of open models argue that access to AI systems helps security researchers respond faster to threats. They have warned that broad restrictions could limit defenders’ ability to analyse and protect against AI-driven attacks.
On July 24, Nvidia joined other companies in signing a letter titled “Open Weights and American AI Leadership”, urging policymakers to support open-weight AI development.
The alliance members are also contributing specific tools and technologies.
Microsoft is providing its MDASH harness, which uses multiple AI agents to identify and verify software vulnerabilities. SpaceX is contributing its Grok Build coding agent and said it plans to release model weights for its Grok AI models.
Hugging Face will offer its Safetensors model weight format to the PyTorch Foundation to support safer storage and use of AI model weights.
The coalition does not include some of the biggest AI model developers, such as OpenAI, Google, Anthropic and Meta, highlighting the divide between companies supporting open AI systems and those focused on closed models.
The launch of Open Secure AI Alliance comes as AI companies invest more in cybersecurity tools designed to detect software weaknesses and protect digital systems. OpenAI and Anthropic have also introduced security-focused AI platforms, placing them in competition with emerging open-source approaches.
Nvidia and its partners built the alliance around the argument that open AI tools can serve as a security resource by giving researchers and organisations more control over how AI systems are tested, monitored and defended.
