Artificial intelligence infrastructure is entering a new phase of development. As demand for GPU computing grows, neocloud providers and AI data center operators are building increasingly sophisticated environments to support organizations running demanding AI workloads. But the technologies enabling this expansion are also creating security challenges that extend beyond traditional cloud and application protection.
Modern AI data centers combine thousands of GPUs, high-speed networking, specialized management controllers, shared storage, and complex orchestration systems. Many of these technologies were originally designed for trusted high-performance computing environments. Their growing deployment in commercial neocloud environments, where unrelated organizations may share physical infrastructure, introduces additional security considerations.
Lava Security is addressing these challenges through its Top 10 Neocloud and AI Data Center Security Risks framework, which has become an official OWASP project. The initiative aims to establish a shared reference for understanding vulnerabilities across the infrastructure powering artificial intelligence.
Developed by Yakir Kadkoda, CTO and Co-Founder of Lava Security, and Michael Katchinskiy, Head of Security Research, the framework brings together research and feedback from more than 20 industry practitioners.
The Infrastructure Layer Requires Greater Attention
One of the defining characteristics of modern neocloud environments is the level of control customers may receive over their rented infrastructure. In bare-metal deployments, customers can receive root access to individual servers, allowing them to manage the underlying operating system and computing resources.
However, this access can create security implications that extend beyond a single customer deployment.
For example, unrestricted access to management interfaces could allow a tenant to modify hardware boot configurations. These changes may survive an operating system reinstall, potentially exposing the next customer if the server is reassigned without resetting the affected settings.
Such scenarios demonstrate why security teams must consider the infrastructure beneath conventional software environments.
“AI infrastructure introduces attack surfaces that traditional security approaches may not adequately address. Organizations need to understand what happens below the operating system, particularly when hardware, firmware, and management interfaces are shared across customers,” said Michael Katchinskiy, Head of Security Research at Lava Security. “This framework helps make those risks easier to identify and provides practical guidance for addressing them.”
Mapping the Risks Across AI Infrastructure
The OWASP Top 10 Neocloud and AI Data Center Security Risks examines vulnerabilities across hardware, firmware, networking, management, and operational layers.
Among its areas of focus are compromised server firmware, vulnerabilities in high-speed networking technologies such as InfiniBand and RDMA, inadequate isolation between tenants, and insecure out-of-band management interfaces.
The framework also addresses supply chain threats, data center facility systems, shared storage and AI artifacts, incomplete security certifications, centralized operational services, and delays in vulnerability patching.
Its approach considers how weaknesses across interconnected infrastructure components can affect the security of entire environments. Each risk category includes potential attack scenarios, real-world implications, and practical mitigation recommendations.
This gives infrastructure providers and operators a resource for evaluating their environments, identifying gaps, and prioritizing improvements. Organizations purchasing GPU cloud capacity can also use the framework to better understand the security controls protecting their workloads.
Establishing a Shared Industry Reference
The framework was initially published in July 2026. Its development drew on disclosed vulnerabilities, incident reports, technical documentation, and Lava Security’s research into AI infrastructure security.
Feedback from more than 20 practitioners also helped shape the initiative. Contributors represented neocloud providers, hardware original equipment manufacturers, security companies, universities, and research institutions.
Moving the framework into OWASP provides an opportunity to expand this collaboration and involve more stakeholders in its development.
“Securing AI infrastructure requires looking at the entire environment, not just individual components. Hardware, networking, management systems, and workloads are interconnected, and risks in one area can affect the security of the broader infrastructure,” said Yakir Kadkoda, CTO and Co-Founder of Lava Security. “Developing this framework within OWASP gives the community an opportunity to contribute its expertise and continuously improve how these risks are understood and addressed.”
Keeping the Framework Relevant
The project will continue evolving through an open development process on GitHub. Security researchers, infrastructure engineers, hardware vendors, and operators can review the framework, challenge existing assumptions, propose changes, and contribute additional research or mitigation recommendations.
Contributors will be credited in future releases.
The initiative is designed to evolve alongside AI infrastructure, including new GPU architectures, networking technologies, management systems, and deployment models.
Interested participants can visit the official OWASP project page or access the GitHub repository to review the framework and contribute to its continued development.
As AI infrastructure becomes more complex, having a shared reference for identifying and addressing its security risks could help providers and users better understand the systems supporting their workloads.
