The rapid expansion of artificial intelligence is creating new demands on the infrastructure supporting it. Thousands of GPUs, high-speed networking systems, shared storage environments, and specialized management technologies are becoming essential components of modern AI operations. But as this infrastructure grows more interconnected, security challenges are emerging in areas that traditional approaches may not fully cover.
One particularly important shift is the growing use of neocloud environments, where multiple organizations share physical infrastructure while maintaining separate workloads. Unlike conventional cloud environments, some bare-metal neocloud deployments give customers root access to rented servers, introducing questions about how hardware, firmware, and management interfaces are secured between tenants.
Lava Security is looking to address that gap through a new industry initiative. The company announced that its Top 10 Neocloud and AI Data Center Security Risks framework has become an official OWASP project, creating an open reference for identifying and mitigating vulnerabilities across AI infrastructure.
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 Security Challenges Beneath AI Workloads
AI infrastructure introduces a different set of considerations from those typically associated with conventional enterprise environments. Modern data centers combine specialized hardware, firmware, networking technologies, orchestration platforms, and management interfaces. A weakness in one layer can have consequences for the broader environment.
The risks become particularly relevant when physical infrastructure is shared between unrelated customers. In bare-metal neocloud environments, customers may receive root access to rented servers. While this provides significant control over their computing environments, it can also create opportunities for attacks that persist beyond an individual operating system installation.
For example, unrestricted access to management interfaces could allow a tenant to modify hardware boot configurations. If those settings survive an operating system reinstall and are not reset before the server is reassigned, a subsequent customer could inherit a compromised environment.
This type of scenario illustrates why infrastructure security cannot stop at the operating system.
“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.”
A Broader View of Infrastructure Security
The OWASP Top 10 Neocloud and AI Data Center Security Risks examines vulnerabilities across several interconnected layers of infrastructure.
Among the areas covered are compromised server firmware, weaknesses in high-speed networking technologies such as InfiniBand and RDMA, inadequate tenant isolation, and insecure out-of-band management interfaces.
The framework also considers supply chain threats, data center facility systems, shared storage and AI artifacts, incomplete security certifications, centralized operational services, and delays in vulnerability patching.
Rather than treating these as isolated technical problems, the initiative examines how weaknesses across different components can affect the security of an entire environment.
Each category includes potential attack scenarios, real-world implications, and practical mitigation recommendations. Infrastructure providers and operators can use the material to evaluate their environments, identify gaps, and prioritize security improvements. Organizations purchasing GPU cloud capacity can also use it to better understand the controls protecting their workloads.
Bringing the Industry Into the Discussion
The framework was initially published in July 2026, drawing on disclosed vulnerabilities, incident reports, technical documentation, and Lava Security’s research into AI infrastructure security.
Its development incorporated feedback from more than 20 practitioners representing neocloud providers, hardware original equipment manufacturers, security companies, universities, and research institutions.
For Kadkoda, the move to OWASP creates an opportunity to make that work more collaborative.
“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.”
An Evolving Security Reference
The framework will continue developing through an open process on GitHub. Security researchers, infrastructure engineers, hardware vendors, and operators can review the project, challenge existing assumptions, propose changes, and contribute additional research and mitigation recommendations.
The initiative is intended to evolve alongside AI infrastructure, including new GPU architectures, networking technologies, management systems, and deployment models.
As organizations increasingly depend on specialized infrastructure to operate AI workloads, establishing a shared understanding of the risks involved could become an important part of securing the systems behind the technology.
The OWASP project provides a starting point for that effort, giving different parts of the infrastructure ecosystem a common reference for identifying and addressing threats that extend beyond conventional software security.
