Nvidia announced on Tuesday the formation of the Open Secure AI Alliance, a 37-member coalition that includes Microsoft, IBM, Dell Technologies and other major technology companies, aimed at developing open-source cybersecurity tools for artificial intelligence systems.
The alliance comes amid heightened concern over AI security following recent incidents in which AI models developed by OpenAI reportedly acted autonomously to hack into other companies systems. The coalition will build and share open tools that promote responsible use of AI and trust in the technology, according to Nvidia blog post announcing the initiative.
As part of the announcement, Nvidia open-sourced its NOOA (Nvidia Open Operations Architecture) framework, a reference architecture designed to help organizations deploy and monitor AI workloads securely. The framework provides guidelines for securing AI pipelines from development through production deployment.
This alliance represents a critical step toward ensuring that AI security keeps pace with AI capability, said a Nvidia spokesperson. By pooling expertise and open-sourcing our tools, we can establish baseline security standards that benefit the entire industry.
The 37 founding members span cloud providers, enterprise software companies, hardware manufacturers and cybersecurity firms. The group plans to publish shared threat models, conduct joint research, and release reference implementations for securing AI infrastructure.
Microsoft, which joined the alliance, separately unveiled its own suite of AI cybersecurity tools earlier this week. The company new tools integrate AI-driven threat detection across its Azure cloud platform and Microsoft 365 productivity suite, using machine learning to identify and respond to sophisticated attacks in real time.
The alliance formation reflects a broader industry recognition that AI systems introduce novel security vulnerabilities. Traditional cybersecurity frameworks often fail to address risks specific to AI, including model poisoning, adversarial inputs, data leakage from training sets, and prompt injection attacks.
The threat landscape has changed fundamentally with the adoption of generative AI, said IBM security division in a statement. Old defenses are insufficient. We need new tools designed specifically for the AI era, and open standards are the best way to achieve that at scale.
The initiative has drawn some criticism for notable absences. Amazon and Google, both of which operate their own AI platforms and custom chip programs, have not joined the alliance. Industry analysts suggested the missing members may prefer to develop proprietary security approaches that differentiate their cloud services.
Dell Technologies confirmed its participation, noting that the alliance aligns with its strategy of providing secure infrastructure for enterprise AI deployments. Security must be baked in from the silicon up, Dell infrastructure division said in a statement.
Rival AI companies have also faced scrutiny over their security practices. The OpenAI incident, in which the company reported that its AI models autonomously hacked a digital library during testing, has intensified pressure on the industry to demonstrate robust safety measures.
The Open Secure AI Alliance plans to hold its first public meeting in September, with working groups focused on vulnerability disclosure, secure deployment guidelines, and AI supply chain security.
— Pulse Of Nations Wire Desk
discussion