- This event has passed.
- There were no results found.
Auditing Large Language Model Ecosystems: From Model Outputs to Agentic System Deployment
October 1, 2025 @ 4:00 pm - 5:30 pm
@
The College of Arts & Sciences’ Ashby Dialogues presents this lecture by Dr. Tianlong Chen, a computer science professor from UNC Chapel-Hill.
The 2025-26 Ashby Dialogues explores the theme, “Rethinking the Algorithm: AI and the Human Experience.”
About this presentation: Large Language Models (LLMs) are rapidly evolving from monolithic entities into complex, interactive ecosystems. While previous research has focused on input-level vulnerabilities, such as adversarial prompts and jailbreaking, a new frontier of pragmatic, system-level risks has emerged.
This presentation shifts the focus from input manipulation to auditing the security and governance of the entire operational stack of LLM agent development. We dissect novel threat aspects that target the ecosystem itself, including knowledge base poisoning, covert prompt injection in agent societies, economic resource exhaustion, and hardware faults. We then survey a new generation of defenses designed to address these unique challenges, ranging from proactive output protection to decentralized monitoring.
Together, these topics provide a blueprint for pragmatic safety, covering both the immediate securing of LLM agentic systems and the principles for their long-term management. Attendees will leave with a comprehensive framework for this next generation of trustworthy AI.
Free and open to the public.
