CrowdStrike Partners with MITRE Center for Threat-Informed Defense to Launch Secure AI Project

This report highlights CrowdStrike’s collaboration with the MITRE Center for Threat-Informed Defense on the Secure AI project to enhance the MITRE ATLAS framework with new AI-focused TTPs (e.g., Acquire Infrastructure: Domains, Erode Database Integrity, Discover/Publish LLM Hallucinations), emerging threats (privacy/membership inference, LLM behavior modification, jailbreaking, tensor steganography), and mitigations (model alignment, guardrails, guidelines, AI Bill of Materials, AI telemetry logging). It emphasizes the growing adversary focus on AI-enabled systems and references a case study involving a large-scale cryptomining incident on Ray clusters, underscoring the need for threat-informed defenses for AI systems.