Crowdstrike Announces New Tools to Protect From AI Cyber Attacks

CrowdStrike has introduced a series of new features for its Falcon cybersecurity platform aimed at protecting enterprises from emerging AI-driven threats. One major addition is AI Model Scanning, designed to detect trojanised AI models—systems that behave normally but produce harmful output when triggered by specific prompts.

These threats can be created by manipulating training data or altering a model’s internal parameters. The tool also identifies backdoors and ranks threats using ExPRT.AI, which assesses exploit risk based on real-world threat intelligence.

“The explosion of AI, combined with multi-cloud complexity, has outpaced patchwork tools that can’t keep up with today’s adversaries, or are limited to single-cloud environments. CrowdStrike is breaking that model. These innovations deliver real-time protection, not just telemetry, across every layer of risk: data, workloads, identities, SaaS, and AI," Elia Zaitsev, chief technology officer, CrowdStrike, said.

Another new feature, the AI Security Dashboard, helps administrators detect “shadow AI,” where employees use unauthorised AI tools. It also monitors the use of internal AI training data.

Falcon Data Protection has also been enhanced. It now detects unauthorised data transfers on Macs and can scan encrypted ZIP files during their creation. It can also identify sensitive data uploads to generative AI tools, even if the data is modified.

Additionally, a new Falcon Data Protection for Cloud version monitors public cloud environments, while a new SaaS Threat Services offering provides expert assessments of enterprise SaaS configurations and integrations.

Powered by the CrowdStrike Security Cloud and world-class AI, the CrowdStrike Falcon platform leverages real-time indicators of attack, threat intelligence, evolving adversary tradecraft, and enriched telemetry from across the enterprise to deliver hyper-accurate detections, automated protection and remediation, elite threat hunting,g and prioritized observability of vulnerabilities.