Irregular is glad to share the publication of a new paper, AI Security Priorities: A Field-Wide Agenda, co-authored with RAND and multiple additional writers from leading organizations. The paper was informed by more than 20 experts from frontier AI labs, industry, government, and academia.
AI systems are being integrated into critical economic, governmental, and national security functions faster than society can adapt. The paper identifies the highest-priority areas for advancing AI security across four themes: establishing strategic foundations and policy frameworks; advancing public-private coordination and institutional infrastructure; advancing technical security engineering and assurance; and governing agentic AI under adversarial pressure.
The ten priority areas with the highest importance
The ten priority areas with the highest cost-effectiveness
The most cost-effective priorities are foundational: shared resources, assessment protocols, and incident response practices. The field currently lacks common tools of this kind. The most important priorities tend to require larger-scale institutional or governmental action: various public-private partnerships between frontier AI labs, security organizations, and government; or deep technical research, such as methods to enable wide adoption of confidential computing. Many of the most important priorities also rank among the hardest to execute, and no single actor can take them all on.
Working with AI labs, enterprises, and government partners, we see the gap between AI capability and AI security widen daily. We hope this paper can empower organizations, individuals, policy makers and others to meaningfully impact the trajectory of AI security.
Read the full paper here.