Israeli cloud security company Upwind has announced a strategic collaboration with NVIDIA to protect the rapidly expanding world of artificial intelligence workloads.
The partnership combines Upwind’s runtime-first security platform with NVIDIA’s advanced AI compute and inference technologies, delivering real-time protection for GPU-powered environments.
Artificial intelligence is transforming industries at an unprecedented pace, powering everything from autonomous agents to large-scale language models.
Yet as enterprises adopt AI infrastructure at scale, they face mounting challenges in safeguarding the integrity, safety, and trustworthiness of their operations against increasingly sophisticated cyber threats.
New attack surfaces are emerging, and traditional security approaches are struggling to keep up.
A press release from Upwind said that the company’s collaboration with NVIDIA was designed to redefine how runtime security is applied to cutting-edge AI environments, using a runtime-first approach and introducing new mitigation techniques against AI-specific cyber risks.
Upwind and NVIDIA partner for AI cloud security
Headquartered in Tel Aviv and founded in 2022 by Amiram Shachar and his partners from Spot.io, which was sold to NetApp for $450 million, Upwind has quickly established itself as a leader in runtime-first cloud security.
The company has raised $180 million from investors, including Greylock, Cyberstarts, Leaders Fund, Craft Ventures, Cerca Partners, and Sheva, a VC fund founded by former NBA player Omri Casspi.
Upwind developed a Cloud-Native Application Protection Platform (CNAPP) that helps organizations focus on real risks by consolidating various security functions into a comprehensive toolkit.
The integration of NVIDIA NIM microservices into Upwind’s platform enhances its internal AI-driven security operations, boosting performance for runtime analytics, vulnerability correlation, and large-scale threat modeling.
In the release, the companies said that the framework they are introducing will give enterprises enhanced performance through: accelerated computing; deployment flexibility across sovereign and private clouds; cost-effective scalability for AI inference and analytics; strict data privacy enforcement; and tailored engineering aligned with customer-specific environments.
At the same time, Upwind now delivers dedicated protection for NVIDIA GPU-based infrastructures, including the NVIDIA DGX platform and the Blackwell architecture, ensuring continuous runtime visibility, risk prioritization, and uncompromised performance.
Upwind has also integrated NVIDIA Garak, an open-source framework for adversarial and robustness testing, into its LLM security validation layer.
Garak simulates prompt injection, jailbreaks, and data exfiltration attempts in large language models.
By combining Garak’s testing capabilities with Upwind’s runtime and API observability, enterprises can continuously validate the safety, integrity, and compliance of their AI applications, ensuring resilience against real-world threats and attempts to manipulate models.
“NVIDIA is setting the foundation for enterprise AI, and we’re proud to both leverage it and secure it,” said Dan Yahav, SVP Platforms at Upwind. “By combining Upwind’s runtime visibility and protection with NVIDIA’s accelerated AI infrastructure, we’re helping organizations deploy AI at scale, safely, efficiently, and with full confidence. Leveraging NVIDIA core technologies such as NVIDIA NIM microservices and NVIDIA Garak, Upwind uses AI to deliver exceptional security outcomes for our customers.”
Ariel Levanon, vice president of cybersecurity at NVIDIA, said that “as AI takes on a central role in business and infrastructure, developers must design systems that are secure from the start. By incorporating NVIDIA’s accelerated computing, advanced AI frameworks, and security-ready infrastructure, Upwind is changing how organizations understand and defend modern cloud environments.”
This collaboration builds on Upwind’s broader AI security strategy, which includes runtime protection for AI workloads, vulnerability management, and LLM-aware API security. Together, Upwind and NVIDIA are setting a new standard for trusted AI, where performance and security advance in lockstep.
NVIDIA’s expanding footprint in Israel
The collaboration comes as NVIDIA is dramatically expanding its presence in Israel, underscoring the country’s importance to its global AI strategy. In 2025 alone, NVIDIA announced a $500 million investment to build one of Israel’s largest AI data center labs at the Mevo Carmel Science and Industry Park.
The company also unveiled plans for a massive R&D campus in the Galilee, spanning up to 180,000 square meters, which could triple its local workforce and create thousands of jobs. These moves build on NVIDIA’s 2019 acquisition of Mellanox Technologies, a deal that transformed its networking and AI capabilities and cemented Israel as a cornerstone of its global operations.