Blog
Meet our Advisors: Bo Li
Professor Bo Li joins ACTAVA as Chief AI Safety and Alignment Advisor. She is an Abbasi Associate Professor at the University of Illinois Urbana-Champaign and co-founder and CEO of Virtue AI, where she builds red-teaming and guardrails for autonomous agents. She explains why healthcare agents have to be secure at the system level.
We are incredibly proud to welcome Professor Bo Li to the actAVA AI team as our newest Strategic Advisor and our Chief AI Safety & Alignment Advisor.
Bo is an Abbasi Associate Professor at the University of Illinois Urbana-Champaign, where her research focuses on trustworthy machine learning, AI security, privacy, robustness, and alignment.
She is also the Co-Founder and CEO of Virtue AII, where she has taken that research into the real world. Bo is building enterprise AI security, red-teaming, guardrails, and governance for models, applications, and increasingly autonomous agents.
That combination of frontier research + enterprise execution is rare.
And it matters deeply to actAVA.
As a leading researcher in trustworthy machine learning and AI security, and a pioneer in evaluating the robustness of large language models, she brings invaluable expertise to our mission to build the ultimate AI orchestration platform for healthcare organizations.
As AI models evolve into autonomous agents, the attack surface in critical industries like healthcare expands beyond simple prompts to encompass complex multi-step actions and clinical workflows. Bo Li's extensive background in securing agentic systems and developing certifiably robust frameworks ensures that actAVA's healthcare AI agents are not only highly capable but structurally secure by design.
"In the era of autonomous healthcare agents, security cannot be an afterthought—it must be foundational. I joined actAVA because they recognize that safety isn't just about protecting individual models; it’s about engineering structurally secure, verifiable, and controllable agentic systems that healthcare organizations can trust."
Why did you decide to be an advisor to actAVA?
My long-term goal, through my work at the Secure Learning Lab (SL2) and as a co-founder of Virtue AI, is to make machine learning systems robust, private, and generalizable with concrete guarantees for real-world applications. actAVA is tackling one of today's most critical challenges: orchestrating agentic AI in policy-dense healthcare environments. In healthcare, one missed policy check or a vulnerability in an AI agent isn't just a technical error—it impacts patient safety. I joined actAVA as an advisor because they recognize that securing individual models isn't enough; AI must be protected at the system level, with security designed in from the start.
What are the biggest changes you see underway in healthcare AI?
We have reached a critical inflection point where AI systems are gaining memory, access to tools, and the ability to act autonomously. In healthcare, this translates to AI agents managing clinical workflows, compliance checks, and operational tasks. The shift from passive prediction to active, agentic execution means the attack surface is rapidly expanding. The biggest change is the industry's realization that we must move beyond standard red-teaming of language models to implementing comprehensive, system-level guardrails, sandboxing agents, and using knowledge-enabled logical reasoning to ensure safe, autonomous execution.
Bonus Round
As a pioneer who developed the "adversarial stop sign" and award-winning DecodingTrust frameworks, what is one thing adversarial attacks have taught you about the limitations of AI?
Adversarial attacks have powerfully reinforced how brittle pure data-driven machine learning can be in the face of unexpected or malicious inputs. Human perception can often contextualize anomalies, but standard AI models can be completely derailed by subtle perturbations. It has taught me that true trustworthiness cannot be achieved through empirical testing alone. We need rigorous, game-theoretic analysis and knowledge-enabled logical reasoning to provide mathematically certifiable robustness, especially when deploying these systems into high-stakes environments like healthcare.
More About Our Advisor
Bo Li is the Abbasi Associate Professor in the Computer Science Department at the University of Illinois at Urbana-Champaign and the co-founder and CEO of Virtue AI. She is on the advisory board of the Center for Artificial Intelligence Innovation (CAII) at Illinois and is a member of the Information Trust Institute (ITI). She is also affiliated with several research centers aiming to broaden research collaboration, such as the Advanced Digital Science Center (ADSC), the Center for Cognitive Computing Systems Research (C3SR), and the Quantum Information Science and Technology Center (IQUIST).
Her research focuses on AI security, trustworthy machine learning, and alignment, with an emphasis on robustness, security, privacy, generalization, and their interconnections. The Secure Learning Lab (SL2) explores different types of adversarial attacks under various constraints and develops robust learning systems for computer vision, natural language processing, safe autonomous driving, and trustworthy agentic systems.
It has to be trustworthy.
It has to be controllable.
And ultimately, it should belong to the customer.
We are incredibly fortunate and excited to have Bo helping us shape this future.
