Jensen Huang Says Cyberattacks Will Grow ‘Exponentially’ — Nvidia, CrowdStrike Build AI ‘Superintelligence’ For Cyber Defense

CrowdStrike and Nvidia announced a partnership to build SafeMind, designed to defend enterprises against increasingly autonomous cyberattacks.

• SafeMind combines offensive and defensive AI models to simulate attacks and continuously improve cyber defenses.
• Nvidia's Nemotron models will be customized for cybersecurity using CrowdStrike's security data and domain expertise.
• Huang said the SafeMind framework could eventually extend beyond enterprise cybersecurity into robotics, factories and other physical AI environments.

Nvidia (NVDA) CEO Jensen Huang on Tuesday said that cybersecurity was at an “inflection point,” warning that “the attacks on companies are going to grow exponentially.” 

Speaking at CrowdStrike’s Fal.Con 2026 conference, Huang called CrowdStrike (CRWD) Nvidia’s “number one cybersecurity partner” and said the two companies were working to give defenders the same advanced AI capabilities increasingly available to attackers.

“We're at an inflection point in cybersecurity for obvious reasons. We have now had agentic AI, the ability to automate attacks, and the attacks on companies are going to grow exponentially.”

– Jensen Huang, CEO, Nvidia

The companies announced a partnership, unveiling a new Cyber Superintelligence Lab called SafeMind, aimed at helping companies defend against increasingly autonomous cyberattacks.

CRWD stock fell nearly 7% in midday trade, while NVDA stock edged over 1% lower amid broader weakness in the market amid rising tensions between the U.S. and Iran as well as the surge in oil prices. 

On Stocktwits, retail sentiment around Nvidia continued to trend in ‘extremely bullish’ territory over the past day, while sentiment around CrowdStrike dipped to ‘bullish’ from the ‘extremely bullish’ zone.

Huang Sees AI Creating New Cybersecurity Arms Race

“I believe in closed models and open models,” Huang stated at Fal.Con. ”We should use closed models and proprietary models and off-the-shelf as much as you can. You know why build something unless you have to.”

However, he added that the cybersecurity industry is moving toward a contest between autonomous AI systems on both sides of an attack.

“Attacks are happening at inference speed,” he said, adding that traditional measures such as breakout time are becoming less relevant as AI agents can perform multiple actions without direct human intervention.

“We don't need every AI to be super smart at everything,” Huang said. “But in some areas we need to be extraordinarily good at something.” According to him, cyber defense is one of those areas.

Huang stated Nvidia's Nemotron models were designed to be customizable and cost-effective for specialized applications such as cybersecurity. He added that CrowdStrike can fine-tune the models using its cybersecurity expertise and data, creating an AI system specifically optimized for detecting and responding to threats.

Nvidia, CrowdStrike Build SafeMind Cyber Defense System

CrowdStrike CEO George Kurtz stated that SafeMind was cybersecurity’s “first complete agentic system” for cybersecurity, combining AI models with a specialized software “harness” within the platform.

SafeMind pairs two models, the first of which is Blue Solano, designed for defensive cybersecurity. The second, called Red Tempest, is designed to simulate sophisticated attacks. The models operate against each other in a digital twin of an enterprise environment, allowing the defensive system to learn from simulated attacks and continuously improve its ability to detect and respond to threats.

Huang said the approach could eventually extend beyond enterprise cybersecurity. He pointed to robotics, factories, edge computing and other physical AI applications as areas where the same framework could be deployed. 

For updates and corrections, email newsroom[at]stocktwits[dot]com.<