CrowdStrike today revealed it is adding additional artificial intelligence (AI) agents into its managed detection and response (MDR) services using a toolkit provided by NVIDIA.
At the same time, CrowdStrike also announced that a Secure-by-Design AI Blueprint built in collaboration with NVIDIA will now be incorporated into NVIDIA OpenShell, an open-source runtime for creating sandboxes where policy-based guardrails for AI agents can be more easily enforced. CrowdStrike is also committed to integrating the Secure-by-Design AI Blueprint across all of its cybersecurity platforms and services.
Announced at the NVIDIA GTC 2026 conference, both initiatives will ultimately help spur additional adoption of AI models developed by NVIDIA to both apply AI agents trained using synthetic data to cybersecurity and protect the thousands of AI agents that an organization might ultimately deploy.
In internal testing, CrowdStrike claims AI agents based on NVIDIA Nemotron Nano and Nemotron Super AI models that have been added to its existing customizable Charlotte AI framework sped up investigations by a factor of five, while improving triage accuracy rates by a factor of three. The company also noted that fine-tuning the NVIDIA Nemotron Nano model achieved 96% higher accuracy in generating investigation queries.
Daniel Bernard, chief business officer at CrowdStrike, said AI agents are, in effect, a new type of digital workforce. In addition to making it simpler for AI agents to automate tasks alongside the human members of a CrowdStrike services team, there is also a pressing need to secure an attack surface that continues to expand as more AI agents are deployed, he added.
Fernando Montenegro, vice president and practice lead for cybersecurity and resilience at the Futurum Group, said the deep collaboration between NVIDIA and cybersecurity vendors such as CrowdStrike clearly shows two sides of the AI and security conversation: one, how organizations are looking to de-risk their AI investments and, secondly, how vendors themselves are using the new capabilities being offered by the AI hardware vendors and frontier labs.
Like it or not, cybersecurity teams are now locked in an AI arms race on two fronts. Adversaries, in addition to embracing AI to launch more sophisticated attacks at higher levels of scale, are also now targeting AI agents that, if compromised, can be used to take control over entire workflows. In effect, in the absence of any meaningful security and governance controls, every time an AI agent is deployed the overall level of risk to business increases.
Unfortunately, many organizations are now deploying AI agents at rates the internal cybersecurity teams are not able to track and assess. As such, the probability that the number of cybersecurity incidents involving AI agents that will occur in the months ahead is only likely to become much higher. That issue is going to prove especially problematic at a time when many organizations during uncertain economic times are not increasing the overall size of their cybersecurity. In fact, some organizations are hoping that investments in AI and automation might even reduce the size of their cybersecurity teams as more tasks are automated.
Hopefully, the number of those incidents will shrink over time as more cybersecurity teams proactively put the controls in place needed to secure and govern AI agents.
In the meantime, cybersecurity teams while hoping for the best might want to prepare for the worst by relying a little more on external expertise in an era where attacks and mitigation efforts are now all occurring at machine speed.
Recent Articles By Author