Lava takes on the AI boom’s data center security blind spot

The artificial intelligence boom is fueling massive investment in data centers, GPU clusters and specialized computing infrastructure. Cybersecurity companies race to protect cloud environments, endpoints and AI applications. Yet the infrastructure powering these technologies remains comparatively underserved.

Lava, an Israeli cybersecurity startup founded by Elay Gliksberg and Yakir Kadkoda, is targeting that gap. The company has developed a security platform for traditional data centers and neocloud providers. It aims to bring modern cloud security capabilities to the infrastructure supporting the AI economy.

“Without data centers, we don’t have AI,” Gliksberg said.

For the founders, securing the infrastructure beneath AI is becoming just as important as protecting the applications running on top of it.

The Infrastructure Security Gap

Modern AI data centers combine GPUs, networking, storage, firmware and management systems. Together they form highly interconnected environments. Neocloud providers add another layer of complexity by offering shared GPU infrastructure to multiple customers. That makes tenant isolation and infrastructure security particularly important.

Lava’s founders believe existing security solutions remain fragmented across these environments. Traditional vendors typically specialize in individual layers. That leaves organizations to piece together findings from network security, vulnerability management, endpoint protection and infrastructure monitoring.

“The thing is, in on-prem, there is no something like that,” Gliksberg said, referring to the attack-path visibility available in cloud security. “Each security vendor is focusing on a different layer, on a different offering.”

Lava wants to connect those separate findings. It wants to show how vulnerabilities and misconfigurations across infrastructure components could combine into exploitable attack paths.

Founders With Deep Infrastructure Experience

Gliksberg and Kadkoda have known each other for approximately 15 years. They first met while serving in a classified Israeli government technology unit. Their professional backgrounds combine infrastructure development, cybersecurity research and experience scaling technology businesses.

Gliksberg previously worked at Bridgecrew, which was acquired by Palo Alto Networks, before joining its Prisma Cloud organization. During his tenure, he helped grow the business from approximately $1 million in annual recurring revenue to $150 million over three and a half years.

Kadkoda spent 11 years in the Israeli military, including four years at the Israel Security Agency. There, he led red teaming and penetration testing efforts involving critical national infrastructure. He later joined Argon Security, acquired by Aqua Security, as director of research.

Their combined experience helped shape Lava’s direction.

“I want to bring the same experience of cloud security, but for on-prem and AI infra,” Kadkoda said.

Connecting Security Across the Stack

Lava’s platform is designed to provide visibility across data center infrastructure. That includes GPUs, networking, storage, firmware, management systems and virtualization environments. Rather than examining each component independently, the company aims to connect security findings. It also aims to identify relationships that could expose an entire environment.

This approach becomes increasingly relevant. AI-powered attacks make it easier to discover and exploit weaknesses across complex systems. A vulnerability in one component can become considerably more dangerous when combined with weaknesses elsewhere in the infrastructure.

“You have to have something like that. It’s a must,” Gliksberg said, describing the need for a more comprehensive approach to identifying infrastructure attack paths.

Defining AI Data Center Security

Beyond its platform, Lava is working to establish a common framework for understanding emerging infrastructure threats. The company helped develop the OWASP Top 10 for Neocloud and AI Data Center Security Risks. The initiative is intended to identify the most important security challenges facing operators of AI infrastructure.

Kadkoda also described an ongoing research program targeting vulnerabilities in neocloud environments. He said Lava researchers had successfully compromised control planes across more than four providers, potentially exposing other customers’ models. The company is also pursuing partnerships with infrastructure manufacturers and neocloud providers.

Securing the Foundation of AI

Lava’s strategy reflects a broader shift in enterprise infrastructure purchasing. Organizations increasingly prefer integrated data center solutions rather than sourcing GPUs, networking and storage independently. This creates an opportunity for security platforms to become part of the initial infrastructure package.

As AI investment accelerates, the founders believe securing the underlying infrastructure will become an increasingly important priority for CIOs, infrastructure executives and CISOs. Lava is betting that the next generation of cybersecurity will need to look beyond applications and cloud workloads. It will need to protect the hardware and interconnected systems making the AI economy possible.