Foreword
Agentic AI—AI systems capable of autonomous decision-making and adaptation—is transforming industries in 2025 and beyond. However, its rise introduces new and complex security risks that the industry is only beginning to understand.
Recognizing this gap, the “Agentic AI Top Threats” initiative was established as the first industry-wide effort to systematically identify security threats in this emerging field. Unlike traditional AI security frameworks, this initiative is independent, ensuring broad accessibility across enterprises, academia, regulatory bodies, and policymakers. This independence allows it to evolve dynamically, integrating diverse expertise to address real-world security concerns.
Importantly, no money was taken from any vendor, nor was any funding given for marketing or other non-technical efforts—this work is entirely voluntary, driven by experts who believe in the need for a shared security foundation for Agentic AI.
Download the document here: Top 10 Agentic AI Security Risks-Key Threats and Mitigation Strategies
This work was initiated by Vishwas Manral, with significant contributions from Ken Huang, Akram Sherif, and Rakshith Aralimatti. Over 50 individuals from 20 leading organizations participated, including:
Precize, where this work began
Cisco Systems
GSK
Palo Alto Networks
Lakera
EY
DistributedApps.ai
Humana
GlobalPayments
TIAA
Meta
DigitalTurbine
HealthEquity
Jacobs
SAP
Sisense
Amazon
As Agentic AI evolves, so will its risks. This document is just the beginning, and future updates will refine and expand on these insights.
For the latest updates and contributions, visit https://github.com/ precize/OWASP-Agentic-AI or email owaspagentic1o@precize.io.
Download the document here: Top 10 Agentic AI Security Risks-Key Threats and Mitigation Strategies
