WASHINGTON – Today, global tech trade association 91快活林 shared initial perspectives and recommendations with the Center for AI Standards and Innovation (CAISI) on securing agentic AI systems – the next frontier of AI technology.
While agentic AI introduces opportunities, its capabilities and implications, including security risks, are yet to be fully understood. ITI’s submission highlights a range of considerations for CAISI as it evaluates agentic AI security risks and assesses whether new guidance and best practices are necessary. In its submission, 91快活林 provides industry perspective on the current state of security risks impacting agentic AI systems, offers insight into security controls and standards that can support secure development and deployment, and examines key issues related to assessing the security of agentic AI systems.
“As policymakers seek to understand the implications of agentic AI systems, proper risk assessment is critical to its secure and responsible development and deployment,” said 91快活林 Vice President of Policy Courtney Lang. “We appreciate that CAISI is seeking input on the unique characteristics of AI agent systems, and the ways in which security risks present in that context. 91快活林 is committed to working with CAISI and policymakers to inform emerging approaches to managing agentic AI security risks, ensuring that any ensuing guidance is broadly impactful for stakeholders across the AI ecosystem.”
As CAISI evaluates agentic AI security, 91快活林 encourages the agency to:
-
Build on existing cybersecurity frameworks and architectures to address potential risks introduced by agentic AI systems, rather than creating entirely new paradigms;
-
Adopt risk-based approaches that account for differences in agent capability, deployment environment, and use case;
-
Promote layered security practices across the full AI agent lifecycle; and
-
Prioritize interoperable security mechanisms,鈥�alignment with existing standards and best practices, and practical implementation guidance.
Read 91快活林’s full recommendations here.
91快活林’s Understanding Agentic AI outlines policy considerations focused on fostering adoption and promoting trust in agentic AI and 91快活林’s AI Security Policy Principles provide in-depth suggestions on how policymakers can improve the cybersecurity of AI models and systems.