AI Security Guide and Risk Assessment Tool
A practical resource for developers, security experts, and policy professionals navigating the AI security landscape. It builds on industry best practices and expert insights to outline threats, vulnerabilities, and prioritized ways to mitigate them.
Images by Quantic69, Jackie Niam, Kindamorphic, Da-Kuk, Piskunov/Getty Images. Design by Haley Okuley/RAND.
AI Security: Why It Matters
Securing artificial intelligence (AI) systems—particularly increasingly powerful, cutting-edge models—is critical. AI is transforming sectors from medicine and scientific research to finance and national defense. As these systems become more capable, interconnected, and embedded in high-impact workflows, their potential security failures grow more consequential. They also become more attractive targets for misuse.
Whether you are training a model, deploying an application programming interface (API), releasing weights, or shaping AI governance, your decisions influence not just performance and usability, but also security, integrity, and risk—to your organization, your users, and society at large.
About the AI Security Guide
This guide is a practical, risk-based resource for developers, security experts, and policy professionals navigating the AI security landscape.1 The guide addresses security of AI systems broadly, including machine learning (ML) models and other AI-enabled architectures. Certain sections, such as the threat landscape and model weight protection sections, focus more specifically on statistical, ML-based models. Building on industry best practices and expert insights, the guide helps you understand and manage the security risks associated with AI systems across their lifecycle—from design and development to deployment and operation.
Our risk-based approach means that this guide and the accompanying tool
- focus on identifying and prioritizing risks specific to your AI system and context
- map proportional, layered security safeguards and controls to each phase of the AI lifecycle
- align with established security standards, such as NIST Special Publication (SP) 800-53 and ISO 27001,2 to support informed, consistent decisionmaking.
The guide does not aim to replace compliance frameworks, legal advice, or expert consultation. Instead, it offers clear, actionable steps for reducing your system’s most relevant risks and building a stronger security posture.
Use the guide and accompanying tool to navigate the evolving AI security landscape with confidence.
- Use the Guide Navigator below to jump straight to the sections most relevant to you, or explore the full guide for a comprehensive overview of AI security risks and defenses.
- Next, use our accompanying Risk Assessment and Security Control Prioritization Tool—an interactive web-based tool that helps you understand your organization’s unique risk profile and then recommends specific, tailored security controls to strengthen your AI system where needed most.
Get Started with the Guide Navigator
Create a custom guide that matches your specific security goals.
Get your customized guideFunding
This project was funded by a grant from the United States Department of State. The opinions, findings, and conclusions stated herein are those of the authors and do not necessarily reflect those of the United States Department of State.
Acknowledgments
The authors wish to thank the numerous people who provided their expertise and insights throughout the project. We thank all the experts who participated in our AI security workshop in June 2025. We thank Ian Mitch for his extensive work on threat modeling and Neeti Pokhriyal, Gaurav Sett, Hiwot Demelash, Joseph Matveyenko, and Morgan Sandler for their support facilitating the workshop. We thank the companies who reviewed and tested the draft guide and tool, and our RAND colleagues Jair Aguirre, Andre Barrera, James Gimbi, Aidan Hesse, Chad Heitzenrater, Wesley Hurd, and Matthew Malone for their testing and feedback on the risk assessment tool. Our colleagues in RAND Europe Mattias Eken, Erik Silfersten, Andrew Gibson, and Andy Skelton had important contributions at multiple stages. John Vahedi, Elisabeth Hammes, and Said Scheck helped with research on AI companies. Nina Ozbardakci, Junyoung Lee, and Melissa Bradley provided critical support in recruiting companies for the pilot. Aidan Hesse, Wesley Hurd, and Joshua Kavner made important contributions to the implementation guidance in the guide and the tool. Ada Ibeanu and Brett Zakheim conducted background research throughout the project. We also thank Don Snyder, Gary Briggs, and Lance Menthe for their thoughtful reviews and recommendations that improved both the guide and tool. Finally, we thank Samantha Anderson, Darrin Magee, Priya Sethi, Raul Avalos, and Daniel Gurley for their guidance and feedback throughout this project.
Notes
- National Institute of Standards and Technology (NIST), “AI Risk Management Framework.” Return to content ⤴
- NIST, Security and Privacy Controls for Information Systems and Organizations; SP 800‑53 Revision 5 Update 1 (Final). International Organization for Standardization (ISO), Information Security, Cybersecurity and Privacy Protection — Information Security Management Systems — Requirements; ISO 27001. Return to content ⤴
Credits
Emily Cantin (Digital Producer), Lee Floyd (Front-end developer), and Haley Okuley (Designer)