AGENT-SEC '26: 1st Workshop on Security, Privacy, and Safety of Agentic AI Systems
Abstract
Agentic systems extend artificial intelligence (AI) models, including large language models and other foundation models, with capabilities such as planning, tool use, persistent memory, external API access, code execution, and collaboration with other agents. These capabilities enable increasingly autonomous operation, but they also create security, privacy, and safety risks that are not adequately captured by model-level analysis alone. The 1st Workshop on Security, Privacy, and Safety of Agentic AI Systems (AGENT-SEC 2026), held in conjunction with the 2026 ACM SIGSAC Conference on Computer and Communications Security (CCS '26), provides a forum for researchers and practitioners to explore and investigate these system-level risks and the corresponding defenses, evaluation methods, and governance mechanisms. The workshop covers attacks on agentic workflows, privacy risks arising from tools and memory, safety and alignment of autonomous agents, secure distributed and federated agents, governance and accountability, real-world applications, and security assurance and benchmarking.