Agentic AI Risk-Management Standards Profile

<span>AI systems that use reasoning to autonomously pursue goals through interaction with external environments and tools, referred to hereafter as "AI agents" or "agentic AI," promise transformative benefits for productivity and complex problem-solving. However, the ability of AI agents to operate with increased autonomy also introduces significant risks, such as unintended goal pursuit, unauthorized privilege escalation or resource acquisition, and other behaviors, such as self-replication or resistance to shutdown, that could result in systemic or catastrophic harm. The unique challenges introduced by agentic capabilities complicate traditional, model-centric risk-management approaches and demand system-level governance that accounts for autonomy, authority, tool access, environments, and interaction effects.&nbsp;</span><span>This paper introduces the Agentic AI Risk-Management Standards Profile ("Agentic AI Profile"), which aims to provide a targeted set of practices and controls for identifying, analyzing, and mitigating risks specific to agentic AI. It is designed to complement the NIST AI Risk Management Framework, function as a specialized extension of the UC Berkeley General-Purpose AI Risk-Management Standards Profile, and address risks that emerge when AI-based systems are granted the agency to act on behalf of users. The Agentic AI Profile also draws on a growing body of technical, policy, and security research on AI agents, autonomy, and AI control.&nbsp;</span><span>It is primarily intended for use by developers and deployers of agentic AI systems, including both single-agent and multi-agent systems built on general-purpose and domain-specific models. Policymakers, evaluators, and regulators can also use the Agentic AI Profile to assess whether agentic AI systems have been designed, evaluated, and deployed in line with leading risk-management practices.</span>