CISA, U.S. and international partners issue guide for secure agentic AI adoption

The Cybersecurity and Infrastructure Security Agency (CISA) and the Australian Signals Directorate’s Australian Cyber Security Centre (ASD ACSC), joined by other U.S. and international partners, have released a joint guide warning that “agentic” artificial intelligence systems—AI tools capable of autonomously taking actions—can introduce new cybersecurity risks even as organizations rush to deploy them for mission-critical tasks.

The document, titled Careful Adoption of Agentic Artificial Intelligence (AI) Services, outlines how these systems can widen an organization’s attack surface, enable unintended privilege escalation, drift from intended behaviors, and obscure incident forensics if not designed and managed with strong safeguards. It is aimed at developers, vendors, and operators in sectors such as critical infrastructure and defense and pairs risk framing with concrete security practices to mitigate emerging threats.

“CISA is committed to supporting the US’s adoption of AI that includes ensuring it aligns with President Trump’s Cyber Strategy for America and is cyber secure,” said CISA Acting Director Nick Andersen. “We actively collaborate with government and international partners on shared priorities with AI advancements while addressing cybersecurity challenges and risks. CISA encourages agentic AI developers, vendors and operators to review this guide.”

Among the actionable steps recommended: enforce least-privilege access and avoid granting broad or unrestricted permissions, especially to sensitive data or critical systems; start with low-risk, non-sensitive use cases to learn and harden defenses; and formally incorporate agentic AI into an organization’s security architecture, governance, and risk management processes.

The guidance reflects mounting interest in agentic AI to streamline operations and decision-making, while acknowledging the higher stakes when autonomous systems interact with tools, data stores, and production environments. It also signals a continued push by cyber authorities to get ahead of risks in safety- and mission-critical settings, following earlier joint advisories on secure AI development and “secure by design” software practices.

Careful Adoption of Agentic Artificial Intelligence (AI) Services is available on CISA’s website and includes best practices, threat considerations, and implementation actions tailored to organizations experimenting with or scaling agentic AI.