The cybersecurity company Check Point Software Technologies collaborates with NVIDIA to deliver an integrated security solution built for AI production systems. AI Cloud Protect is now available for on-premises enterprise environments to protect the development of AI models, agent-based AI applications and inference workloads without any impact on performance. The solution has been verified on NVIDIA RTX PRO Servers, enabling enterprises to securely deploy AI – from data center to cloud – based on NVIDIA BlueField technology.

“As companies race to build AI-driven innovation, they can’t afford to have blind spots,” said Nataly Kremer, product manager at Check Point. “Together with NVIDIA, we’re securing AI factories from the ground up – protecting models, data, and infrastructure without stifling innovation.”.

Today's AI systems face unprecedented security challenges. More than half of all enterprise networks now use AI tools, making them attractive targets for cyberattacks. Data from Check Point shows that 1 in 80 GenAI prompts expose sensitive information. Meanwhile, a new report from Gardener that 32 percent of organizations have been exposed to AI attacks involving prompt manipulation, and 29 percent have suffered attacks on their GenAI infrastructure in the past year. As companies are rushing to adopt AI It has become urgent to protect the entire AI chain – from development to production.

Advertisement

AI Cloud Protect secures AI factories by protecting the cloud infrastructure that powers the development, training, and commercial deployment of AI models. Powered by NVIDIA BlueField-3, AI Cloud Protect delivers a comprehensive security layer without consuming CPU or GPU resources, eliminating the performance penalty that traditional security solutions often incur. With network-based protection, AI Cloud Protect secures AI infrastructure from unauthorized access and reduces the risk of model poisoning (data poisoning) and model support (model exfiltration). At the system level, AI Cloud Protect uses the unique direct memory access provided by NVIDIA DOCA Argus to provide complete visibility into all processes running on each AI node. This enables Check Point to detect and prevent malicious processes and workloads at the host level, even those hiding in publicly downloaded Large Language Models (LLMs).

“Security is critical for next-generation AI infrastructure,” said David Reber, chief security officer at NVIDIA. NVIDIA is working with Check Point to integrate BlueField acceleration and the NVIDIA DOCA Argus runtime security framework into the AI Cloud Protect platform to help enterprises confidently deploy AI.

AI Cloud Protect will also support the upcoming NVIDIA BlueField-4 The DPU, which delivers up to six times the compute capacity and twice the network speed for AI inference and data warehousing processes, enables scalable and secure AI operations with unmatched performance.