Panel Discussion

As enterprise AI agents move from static deployments to continuously adapting systems, the research community faces an expanding agenda of interconnected challenges. This panel brings together perspectives from academia, industry, and policy to chart that agenda across several fronts: What conceptual frameworks best capture the unique dynamics of continual learning in enterprise settings? How should we design benchmarks that evaluate not only learning performance but also safety drift and adversarial robustness over time? What role can open-source projects and shared evaluation infrastructure play in accelerating progress? How do we enable meaningful data sharing and collaboration without compromising privacy or trust? And critically, how do we embed security and governance — from regulatory compliance to alignment stability — as first-class concerns throughout the continual learning lifecycle rather than afterthoughts? Panelists will explore these questions and work toward a community-wide research agenda for building enterprise AI agents that remain capable, safe, and accountable as they evolve.