CISA and International Partners Publish Guidance on Secure Agentic AI Adoption

Cybersecurity and Infrastructure Security Agency (CISA), together with the Australian Signals Directorate’s Australian Cyber Security Centre and other international partners, announced in a press release the publication of a joint guide titled Careful Adoption of Agentic Artificial Intelligence (AI) Services.

The guide addresses cybersecurity challenges linked to the introduction of agentic AI systems, which are increasingly used in critical infrastructure and defense operations. It outlines potential risks such as expanded attack surfaces, privilege escalation, behavioral misalignment, and incomplete event records.

CISA Acting Director Nick Andersen stated that the agency supports AI adoption aligned with national cybersecurity strategies and continues to collaborate with international partners on AI security priorities. The document provides best practices for developers, vendors, and operators to secure agentic AI systems.

Recommendations for organizations include limiting access to sensitive data, starting with low-risk use cases, and integrating agentic AI security into existing security models and risk assessments.