LAS VEGAS: Cyberattacks are becoming faster and increasingly automated, and cybersecurity companies say defence must evolve just as quickly.

At CrowdStrike’s Fal.Con 2026 conference in Las Vegas, NVIDIA founder and CEO Jensen Huang joined CrowdStrike founder and CEO George Kurtz to announce CrowdStrike SafeMind, a new AI-powered cybersecurity system designed to help organisations detect, respond to and defend against increasingly sophisticated attacks.

“We’re at an inflection point in cybersecurity,” Huang told an audience of about 10,000 security professionals. “Attacks are now automated. Defense has to be, too.”

SafeMind combines CrowdStrike’s cybersecurity expertise and threat data with NVIDIA’s Nemotron open AI models. The system uses AI agents to continuously test defensive systems against potential attacks, allowing the technology to learn and strengthen its response over time.

CrowdStrike said SafeMind is designed to bring offensive and defensive AI together in a continuous cycle. In simple terms, the system allows AI to simulate attacks, identify weaknesses and improve its defences before real attackers can exploit them.

Kurtz said the development addresses a growing imbalance in cybersecurity, where attackers have increasingly gained access to advanced AI tools.

“The real gap that I saw was that the attackers had frontier AI, and the defenders didn’t,” Kurtz said. “And that changes now.”

CrowdStrike said its internal testing found that one of its models, developed using NVIDIA’s Nemotron 3 Super and trained with CrowdStrike’s cybersecurity data, achieved higher accuracy than leading frontier models at significantly lower cost.

The company said its security teams were able to train the models using its own threat data without sending that information to an external AI provider. This gives organisations greater control over the data and models used to protect their systems.

Huang described the relationship between an AI model and its security tools as similar to a brain and an exoskeleton. The AI model provides the intelligence, while specialised software tools turn that intelligence into an autonomous cybersecurity agent.

CrowdStrike also announced Falcon IQ, which is designed to automate cybersecurity workloads through AI agents, as well as an expanded version of its Guardian AI safety solution.

The announcements come as AI-driven cyberattacks accelerate. CrowdStrike said AI-enabled attacks increased 89% over the past year, while the fastest recorded breakout time for an eCrime attack has fallen to just 27 seconds.

For businesses, the message is becoming increasingly clear: traditional, human-speed cybersecurity may no longer be enough.

As Huang put it, the next stage of cybersecurity will depend on giving defenders AI capabilities that can operate at the same speed — or faster — than the attackers they are trying to stop.

Continue reading — https://blogs.nvidia.com/blog/nvidia-crowdstrike-fal-con-2026/