OWASP tackles AI risk in bold new push

OWASP’s GenAI Security Project introduced three AI security resources—the Top 10 for Agentic Applications 2026, a 250-page AI Testing Guide aligned with the EU AI Act and NIST AI RMF, and the AI Vulnerability Scoring System (AIVSS)—to help organizations identify, test, and prioritize risks in agentic AI systems. The report highlights key risks such as agent goal hijack, tool misuse, identity and privilege abuse, memory/context injection, and cascading failures, while offering a standardized testing approach and a scoring model to quantify AI risk. Expert commentary underscores the value of a shared baseline and tooling, but also cautions against checklist mentalities and over-scoring, emphasizing the need to focus on access control and real-world prioritization.