Nvidia Corp. today introduced DOCA Argus, a new cybersecurity framework designed to protect enterprise artificial intelligence infrastructure at runtime.

Built on Nvidia’s BlueField data processing units, DOCA Argus enables agentless, real-time threat detection that integrates with existing enterprise security systems without affecting performance. It does so by operating on every node to detect and respond to attacks immediately on AI workloads, integrating with enterprise security systems to deliver instant threat insights.

Under the hood, the DOCA Argus framework provides runtime threat detection by using advanced memory forensics to monitor threats in real time. The result is an ability to deliver detection speeds up to 1,000 times faster than existing agentless solutions without affecting system performance.

The speed comes from the way Argus runs. Unlike conventional tools, Argus runs independently of the host and requires no agents, integration or reliance on host-based resources. Nvidia says the agentless, zero-overhead design enhances system efficiency and ensures resilient security in any AI compute environment, including containerized and multitenant infrastructures.

Additionally, by operating outside the host, Argus remains invisible to attackers, even in the event of a system compromise.

The framework can be used by cybersecurity professionals with existing security information and event management, security orchestration, automation and response, and extended detection and response security platforms. That enables continuous monitoring and automated threat mitigation, expanding existing cybersecurity capabilities for AI infrastructure.

“Cyber defenders need robust tools to effectively protect AI factories, which serve as the foundation for agentic reasoning,” said David Reber, chief security officer at Nvidia. “The DOCA Argus framework enables real-time security insights, ushering in a new era of autonomous detection and response — delivering actionable intelligence and giving defenders a data advantage.”

Cisco collaboration

Nvidia also today announced a deepening collaboration with Cisco Systems Inc. to deliver the Cisco Secure AI Factory with Nvidia, a jointly developed architecture that brings built-in cybersecurity to every layer of enterprise AI infrastructure.

The expanded partnership integrates BlueField DPUs and the new DOCA Argus framework into Cisco’s infrastructure solutions to enable real-time, agentless threat detection and secure data processing directly at the hardware level. The result is a scalable solution that embeds runtime protection into AI workflows from the outset rather than bolting on security post-deployment.

“Now is the time for enterprises to be driving forward with AI, but the key to unlocking innovative use cases and enabling broad adoption is safety and security,” said Jeetu Patel, executive vice president and chief product officer at Cisco. “Nvidia and Cisco are providing enterprises with the infrastructure they need to confidently scale AI while safeguarding their most valuable data.”

Agentic AI in cyber

Alongside the new launch of DOCA Argus, Nvidia also took the opportunity to spotlight its growing focus on agentic AI security, outlining a comprehensive framework to protect AI agents that can perceive, reason and act autonomously.

The agents are increasingly used in cybersecurity operations to accelerate threat detection, triage alerts and assist analysts, but they also introduce novel risks. Nvidia believes that there is a need for a dual strategy, defending with agentic AI while also defending against it, to ensure safe deployment in enterprise environments.

To support this, Nvidia has developed an integrated stack of tools and frameworks designed for every stage of the agent lifecycle. Tools include red-teaming and pre-deployment testing using Garak, runtime behavior control through NeMo Guardrails, and confidential computing technologies that protect data in use across single and multi-GPU systems.

Additional tools like the AI-Q blueprint and the Nvidia Agent Intelligence toolkit allow enterprises to rapidly build and scale agentic systems with built-in governance and safety protocols. Major partners, including CrowdStrike Holdings Inc., Trend Micro Inc. and Amdocs Ltd., are already integrating these technologies to enhance the resilience and efficiency of their cybersecurity operations.

Nvidia also highlighted the importance of securing the underlying infrastructure powering these agents. BlueField DPUs and the DOCA AppShield framework provide memory-level threat detection and policy enforcement without relying on host CPUs, reducing the blast radius in case of compromise.

The tools are part of a zero-trust architecture that ensures AI agents operate safely within enterprise workflows, even as they gain more autonomy. As agentic AI expands across industries and critical infrastructure, Nvidia is aiming to make trust, control and security core components of every AI deployment.

Image: Nvidia