OWASP：2026年AI智能体安全与治理现状报告（英文版）

v1.0 of this report (July 2025)framed agentic risks as a portfolio of plausible threats, surveyed an early ecosystem, and called for governance to keep pace. In the year since, adoption has accelerated, the OWASPTop 10 for Agentic Applications has been published, and a regulatory environment around agent-driven harm has begun to take shape. v2 reads that year of evidence and is organized around three findings.