CrowdStrike Uses On-Device AI to Spot Sensitive Data in Real Time
Protecting sensitive business data isn't just about spotting obvious patterns like credit card or tax file numbers. A lot of sensitive information, such as credentials shared in a chat message or details buried in support tickets and documents, doesn't follow a predictable format. Traditional rule-based security tools struggle to catch this kind of unstructured data because there's no fixed pattern to search for.
CrowdStrike has addressed this by building AI language models into its Falcon Data Security product that run directly on the device itself, rather than in the cloud. Working with Intel, the company designed the system to use dedicated AI hardware (Intel's NPU) so models can understand context and meaning in real time. This matters because sending sensitive data to the cloud for analysis introduces both delays and privacy risks, and running these models on a normal laptop CPU was too slow to keep up with real-time protection needs.
This on-device approach is described as the first step in a wider plan to expand AI-based data classification across different types of hardware, building on CrowdStrike's existing rule-based detection systems.