The expansion of artificial intelligence is creating a new infrastructure ecosystem, one that depends on specialized computing environments and increasingly interconnected technologies. Neocloud providers, AI data center operators, hardware manufacturers, and organizations running AI workloads all have a stake in keeping these environments secure.
Yet the security challenges are not necessarily confined to the systems that organizations interact with directly. Hardware configurations, firmware, networking technologies, management interfaces, and operational services can all introduce vulnerabilities that affect the broader infrastructure.
Lava Security is bringing attention to these issues through its Top 10 Neocloud and AI Data Center Security Risks framework, which has become an official OWASP project. The transition creates an open initiative where security researchers and infrastructure practitioners can contribute to a shared understanding of the threats facing AI infrastructure.
The framework was developed by Yakir Kadkoda, CTO and Co-Founder of Lava Security, and Michael Katchinskiy, Head of Security Research.
An Ecosystem With Shared Security Responsibilities
The infrastructure supporting AI workloads is built around multiple interconnected technologies. Modern AI data centers combine thousands of GPUs with high-speed networking, specialized management controllers, shared storage, and complex orchestration systems.
Many of these technologies originated in trusted high-performance computing environments. Their growing deployment in commercial neocloud environments, where unrelated organizations may share physical infrastructure, creates additional security considerations.
For example, customers in bare-metal neocloud environments may receive root access to rented servers. If unrestricted management interfaces allow them to modify hardware boot configurations, those changes could persist after an operating system reinstall.
Without resetting those configurations before reassigning the server, a provider could inadvertently expose a subsequent customer to a compromised environment.
The example highlights how infrastructure security depends on coordination between providers, operators, hardware manufacturers, and customers.
“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 Framework That Connects Different Security Domains
The OWASP Top 10 Neocloud and AI Data Center Security Risks examines vulnerabilities throughout the infrastructure stack.
Its scope includes compromised server firmware, vulnerabilities in InfiniBand and RDMA networking technologies, inadequate tenant isolation, and insecure out-of-band management interfaces.
It also covers supply chain threats, data center facility systems, shared storage and AI artifacts, incomplete security certifications, centralized operational services, and delays in vulnerability patching.
The framework is designed to connect these different risk areas rather than treat them as independent technical concerns. Each category includes potential attack scenarios, real-world implications, and practical mitigation recommendations.
For providers and operators, this offers a reference for evaluating infrastructure security and prioritizing improvements. Organizations purchasing GPU cloud capacity can also use it to better understand the security controls protecting their workloads.
Turning Research Into Collective Progress
The framework was first published in July 2026, drawing on disclosed vulnerabilities, incident reports, technical documentation, and Lava Security’s research into AI infrastructure security.
Its development also included feedback from more than 20 practitioners across neocloud providers, hardware OEMs, security companies, universities, and research institutions.
The move to OWASP gives the initiative an open development structure through which additional stakeholders can contribute their knowledge and challenge existing assumptions.
“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 Open Invitation to Infrastructure and Security Practitioners
The initiative will continue evolving through an open process on GitHub. Security researchers, infrastructure engineers, hardware vendors, and operators can review its content, challenge existing assumptions, propose changes, and contribute additional research or mitigation recommendations.
The project also provides contributor recognition, with participants credited in future releases.
Its ongoing development is intended to reflect changes in the infrastructure landscape, including new GPU architectures, networking technologies, management systems, and deployment models.
For practitioners interested in the project’s development, the official OWASP project page provides information about the initiative, while the GitHub repository offers a channel for reviewing its contents and contributing to future iterations.
By placing the framework within OWASP, Lava Security is helping create an environment where infrastructure security research can be examined and developed collectively. As AI data centers become increasingly important to the organizations operating AI workloads, that kind of shared effort can provide a useful foundation for understanding and addressing the risks beneath the technology.
