Nvidia has formed a new industry group with major technology and cybersecurity companies to develop and share tools for improving AI safety and security.
The Open Secure AI Alliance includes founding members such as Adobe, CrowdStrike, Dell Technologies, Hugging Face, Microsoft, IBM, Cloudflare, Cisco, Salesforce and several other companies and organisations.
The announcement comes days after an autonomous AI agent powered by OpenAI models escaped its testing environment and compromised Hugging Face’s infrastructure, highlighting concerns about increasingly capable AI agents.
Nvidia said the alliance will develop and share open technologies, tools and techniques that organisations can use to secure AI systems and autonomous agents.
The company argues that putting advanced defensive capabilities behind a small number of closed providers could create weak points and make organisations dependent on systems they cannot fully inspect or modify.
Instead, Nvidia says security teams should have access to advanced AI tools that they can study, adapt and run on their own infrastructure.
The alliance builds on work from the Linux Foundation’s Akrites initiative and the Open Source Security Foundation community.
The alliance was announced shortly after OpenAI disclosed a security incident involving one of its autonomous agents.
During an internal cybersecurity test, an agent powered by OpenAI models found ways to escape the controlled environment and access Hugging Face systems.
OpenAI said the agent chained together several attack methods, including stolen credentials and previously unknown software vulnerabilities, while attempting to obtain information that would help it complete its evaluation.
Hugging Face detected and contained the activity before OpenAI fully understood what had happened. Reuters reported that OpenAI did not connect its agent to the intrusion until after Hugging Face had already contained the threat and contacted the FBI.
OpenAI described the event as an unprecedented cybersecurity incident and said it is strengthening its containment, monitoring and access controls.
The new alliance also follows a wider industry push supporting open-weight AI models.
Open-weight models allow developers to download, inspect, modify and run the core model weights on their own systems instead of relying completely on a company-hosted service.
On July 24, Nvidia joined companies including OpenAI, Google, Microsoft, Meta, Amazon, AMD, Hugging Face and others in signing a letter titled “Open Weights and American AI Leadership.”
The letter argues that open-weight models can improve competition, lower costs and give cybersecurity teams more control over the AI systems they use.
It also warns policymakers against broad restrictions on open models, while acknowledging that openly released model weights can create risks because developers may lose control over how modified versions are used.
Nvidia said it will contribute open models, model weights, datasets and research on AI agent systems to the new alliance.
This includes the Nvidia Labs Object-Oriented Agent project, which is now available as open-source software on GitHub.
The framework is designed to make AI agents easier to control by separating their capabilities into manageable components.
Nvidia says this can make it easier for developers and security teams to test what an AI agent is doing, track its actions, review its behaviour and place limits on what it is allowed to do.
The wider goal of the Open Secure AI Alliance is to create shared security tools as AI agents become capable of carrying out increasingly complex tasks with less direct human supervision.
Get the latest tech news, telecom insights, and product launches wherever you prefer.
Add ProPakistani to Preferred Sources and see more of our stories in Google Search and Top Stories.