Check Point Software Technologies has partnered with NVIDIA to introduce AI Cloud Protect, a comprehensive solution designed to protect so-called “AI factories.” The tool is now available for on-premises enterprise environments and promises to shield the entire AI lifecycle without impacting system performance.
The solution covers everything from the model development phase to autonomous applications and inference loads, ensuring that companies can operate safely in complex AI environments. AI Cloud Protect has been validated on NVIDIA RTX PRO Servers, allowing you to deploy AI with confidence, both in data centers and in the cloud, leveraging NVIDIA BlueField technology.
“As companies accelerate their AI-powered innovation, they can’t afford blind zones. With NVIDIA we are making AI factories secure by design: we protect models, data and infrastructure without slowing down innovation,” explains Nataly Kremer, Chief Product Officer at Check Point Software.
Increasing risks in AI environments
AI systems face increasingly sophisticated threats. More than half of business networks use artificial intelligence tools, becoming priority targets for cybercriminals. Recent data from Check Point Software reveals that one in eighty generative AI prompts exposes sensitive information. Additionally, a Gartner report indicates that 32% of organizations suffered attacks related to prompt manipulation and 29% were victims of cyberattacks targeting their generative AI infrastructure in the last year.
In this context, protecting the entire AI development chain, from creation to production, has become a critical need. The lack of security can lead to data leaks, model corruption and intellectual property risks, directly affecting innovation and business competitiveness.
Protecting the entire AI development chain, from creation to production, has become a critical need
Comprehensive security across the entire AI chain
AI Cloud Protect protects the cloud infrastructure that supports the development, training, and deployment of AI models. Running on NVIDIA BlueField-3 DPU, the solution offers full-stack security without consuming CPU or GPU resources, avoiding the performance compromises common in traditional tools.
At the network level, the platform prevents unauthorized access and reduces the risk of model exfiltration. On the host, it uses NVIDIA DOCA Argus technology to monitor all active processes, blocking malicious activity and hostile workloads, including large publicly downloaded language models.
“Security is essential for the next generation of AI infrastructure,” said David Reber, Chief Security Officer at NVIDIA. “NVIDIA is collaborating with Check Point Software to integrate BlueField acceleration and the NVIDIA DOCA Argus runtime security framework into AI Cloud Protect, helping businesses deploy AI with complete confidence.” The platform will soon support NVIDIA BlueField-4 DPU, which will multiply computing power and network performance, enabling secure and scalable AI operations.
Beyond infrastructure
Check Point Software not only offers infrastructure-level protection, but also secures applications and users. CloudGuard Web Application Firewall, with built-in Lakera capabilities, protects AI applications from threats such as prompt injection and jailbreaking. Infinity GenAI Protect facilitates the safe adoption of generative AI by detecting used applications, enforcing security policies, and ensuring regulatory compliance.
AI Cloud Protect pilot projects are being implemented in AI data centers in the financial sector and with strategic partners such as World Wide Technology. Chris Konrad, Global Vice President of Cybersecurity at WWT, comments, “As organizations build large-scale AI server factories, the combination of AI Cloud Protect and NVIDIA BlueField acceleration delivers enterprise-grade protection for sensitive workloads, from model training to inference, without compromising the performance that modern AI applications demand.”
Security challenges and solutions
AI factories present large attack surfaces due to the massive flow of data and computing power. Emerging threats include prompt injection, jailbreaking, and model poisoning. There is also a high risk of information leaks resulting from the uncontrolled use of AI tools.
AI Cloud Protect addresses these challenges with comprehensive network, host, and application protection, accelerated security with no performance impact, and unified management to orchestrate security across thousands of AI nodes. This way, organizations can reliably innovate with AI, ensuring comprehensive protection of their data and models at all stages of the lifecycle.
