Report: Attackers using AI as weapon to target enterprise AI tools

A new report from CrowdStrike reveals that attackers are increasingly using AI as a weapon to target enterprise AI tools. The research highlights how threat actors are exploiting AI systems through prompt injection, model manipulation, and data poisoning attacks. The report analyzes over 1,200 AI-related security incidents from the first half of 2026, finding a 340% increase compared to the same period in 2025. Key findings include: 78% of organizations using AI tools reported at least one AI-related security incident; prompt injection attacks accounted for 42% of all AI security incidents; data poisoning attacks targeting ML training pipelines increased 580% year-over-year. The report also identifies emerging attack vectors including adversarial inputs designed to bypass AI safety guardrails, model extraction attacks that steal proprietary AI model weights, and supply chain attacks targeting open-source AI libraries. CrowdStrike recommends organizations implement AI security governance frameworks, deploy AI-aware threat detection systems, and conduct regular red team exercises focused on AI attack scenarios.