CoreWeave and CrowdStrike are rethinking AI-native cybersecurity

AI-native cybersecurity requires defenses that can keep pace with attackers while protecting infrastructure that never stops running. That challenge is to bring security expertise and specialized computing closer together and to integrate automation into all the systems companies use to train and run models.

According to Bartley Richardson (pictured left), head of AI at CrowdStrike Holdings Inc., the fastest observed breakout time of eCrime attackers in 2025 was 27 seconds, underscoring the urgency of a faster response. To meet this challenge, computing capacity is required for both training and inference.

“People can’t fight this alone,” he said. “You need machine-speed defense, and that requires critical infrastructure.”

Richardson and Jim Higgins (right), Chief Information Security Officer of CoreWeave Inc., spoke with Dave Vellante and John Furrier of theCUBE Research at the Fully Connected event during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed machine-fast defense, infrastructure-level security and operational trust. (*Disclosure below.)

Cyber ​​models turn attacks into training data

The partnership combines CrowdStrike’s security data and adversary expertise with CoreWeave’s infrastructure capabilities. According to the company, CoreWeave provides training and inference capabilities for SafeMind and serves as a design partner for CrowdStrike’s Cyber ​​Superintelligence Lab.

“The key to the game … is adversarial co-evolution, which is a fancy way of saying, ‘We train and control the best offensive capabilities to attack and find ways to circumvent defensive operations,'” Richardson said. “But then we train these defense models on large infrastructure to learn from these attacks.”

CoreWeave needs to strengthen security while keeping its production systems running. To do this, it is necessary to integrate protective measures into ongoing operations, says Higgins.

“This is what I call the Aikido of the security problem,” he said. “Anything I can do to influence safety while the machine is running and use its own energy to promote safety requirements is exactly what I do. Using AI to assist and scan my code is incredibly important, but so is being able to.” [continuous integration and continuous delivery] Pipeline to use automation to fix misconfigurations.”

AI-native cybersecurity still starts with the basics

Attackers stick to budgets and timelines, making the easiest route into an environment an attractive target. According to Higgins, the combination of several common weaknesses can give them this opportunity.

“If you say something is safe, test it and test it often to be sure, but also know the basics,” he said. “Patch your stuff and continually test and monitor these issues.”

CrowdStrike invests in automation to support continuous security operations. According to Richardson, these capabilities must earn the trust of both CoreWeave and its customers.

“People ask, ‘When will we reach the end of AI?’ And I say, ‘Never,'” he said. “And then I tell them, ‘That’s because it’s AI until you trust it. Then you start thinking of it as automation.’ As we invest in always-on AI, how trust is built into the system is so important to us.”

Here is the full video interview, part of SiliconANGLE and theCUBE’s coverage of the Fully Connected event:

(*Disclosure: TheCUBE is a paid media partner for the Fully Connected event. Neither CoreWeave, the sponsor of theCUBE’s coverage, nor other sponsors have editorial control over theCUBE or SiliconANGLE content.)

Photo: SiliconANGLE