New resource defines 70 key terms across deployment, governance, security, infrastructure and technical concepts to support informed policymaking
SAN JOSE, Calif. — June 16, 2026 — The Silicon Valley Leadership Group (SVLG) today released a new Agentic AI Glossary designed to help policymakers, business leaders, civic stakeholders and other interested audiences better understand one of the most important emerging areas of artificial intelligence.
As organizations move from experimenting with generative AI tools to integrating AI into real workflows, agentic AI represents a significant next step: systems that can interpret goals, plan across multiple steps, use tools, and support task completion across digital and, increasingly, physical environments. The glossary gives nontechnical leaders a shared vocabulary for understanding how these systems work, where they may create value, and what kinds of oversight, security and accountability questions they raise.
“California is home to the world’s leading AI companies, and we have a responsibility to help ensure policymakers and civic leaders understand the technologies that are beginning to reshape our economy and public institutions,” said Ahmad Thomas, CEO of the Silicon Valley Leadership Group. “Agentic AI has the potential to improve productivity, strengthen services and expand human potential, but good policy starts with clear understanding. This glossary is meant to help bridge that gap.”
The glossary defines 70 terms across seven business- and policy-facing categories: core concepts; how agentic systems work; enterprise deployment and business value; governance, risk and accountability; security and failure modes; infrastructure, standards and interoperability; and technical reference material.
Among the key terms defined in the glossary:
- Agentic AI: AI systems designed to pursue a goal, make decisions, use tools and complete multi-step tasks with varying degrees of human oversight.
- Autonomy Spectrum: A framework for describing how much independence an AI system has, from manual human control to supervised autonomy to highly autonomous operation.
- Human-in-the-Loop: A design approach in which a human must review, approve or guide an agent before it takes certain actions.
- Traceability: The ability to reconstruct what an agent did, what information it used, which tools it called, what decisions it made and where human approvals occurred.
- Prompt Injection: An attack in which malicious or conflicting instructions are inserted into content the agent reads, potentially causing the system to act on those instructions through connected tools.
- Agent Observability: The ability to monitor an agent’s behavior, tool calls, failures, costs, latency and escalation points.
“Agentic AI should not be understood simply as a more advanced chatbot,” said Ziyang Fan, Founding Executive Director of the Institute for California AI Policy at the Silicon Valley Leadership Group. “These systems can combine models, tools, memory, planning and oversight to complete tasks across workflows. Policymakers need a practical vocabulary to understand both the opportunity and the risk.”
Unlike traditional software automation, agentic AI systems may be able to interpret a goal, break it into smaller tasks, call external systems, review results and continue working until a task is complete or a checkpoint is reached. That capability can open new opportunities in areas such as customer service, cybersecurity, research, software development, compliance, infrastructure and public-sector modernization. It also raises important questions about authority, human review, data access, reversibility, liability, security and system accountability.
The glossary was developed through SVLG’s Institute for California AI Policy (ICAP), with input from SVLG members participating in the organization’s Agentic AI task force and related workshops.
For policymakers, the glossary is designed to support more precise conversations about AI adoption and governance. Clear terminology can help decision-makers distinguish between different levels of autonomy, identify appropriate use cases, understand when human oversight is needed, and evaluate whether existing laws, procurement rules and risk-management frameworks are prepared for more action-oriented AI systems.
“Agentic AI is moving quickly from concept to implementation in the public sector,” Fan said. “The goal of this glossary is to make sure policymakers and stakeholders have the language they need to ask better questions, design better safeguards and support innovation that delivers real public benefit.”
SVLG released the glossary as part of its broader work to help California move from AI innovation leadership to practical implementation leadership. By equipping policymakers and stakeholders with clearer language, SVLG aims to support a more informed public conversation about how emerging AI systems can advance economic competitiveness, strengthen services, support workers and benefit communities.
About the Silicon Valley Leadership Group
The Silicon Valley Leadership Group is a business association representing many of the world’s most innovative companies and working to advance policies that strengthen Silicon Valley’s economy, quality of life and competitiveness. SVLG partners with public, private and civic leaders to address the region’s most pressing challenges and ensure California remains a global leader in innovation.
About the Institute for California AI Policy
The Institute for California AI Policy is an initiative housed at the Silicon Valley Leadership Group that works at the intersection of technology, policy and public impact to help California lead in the responsible development and deployment of artificial intelligence. The Institute advances research, policy analysis and cross-sector collaboration to ensure AI strengthens California’s competitiveness, communities and public institutions.
###