As artificial intelligence moves beyond simple chatbots toward fully fledged work agents, questions of accessibility and security are becoming increasingly urgent. Speaking at the World AI Conference in Shanghai this month, Li Jingqiu, a product architect at Chinese internet giant Baidu, illustrated that dual challenge with a striking example: two rural teachers in their fifties, with no technical training or experience in coding or AI, built a working content‑selection system in under an hour for a local reading initiative.
Accessibility: AI agents within reach of the general public
Li’s anecdote shows how AI agents can lower technical barriers and enable non‑specialists to solve problems that once required expert skills. This progress points to a real democratization of automation tools—especially for pragmatic, local use cases, such as rapidly deploying a system to curate educational content.
Security and control of AI agents
That accessibility comes with risks. Li warned that as AI agents perform tasks autonomously, ensuring their safety and controllability becomes critical. She outlined practical measures: isolate automated actions to limit their scope and restrict high‑risk operations to prevent unwanted consequences. These safeguards are intended to preserve human oversight over increasingly powerful decision‑making capabilities.
Diverse responses across regions
Li also noted that responses and approaches vary by region—governance, regulatory frameworks, and local adoption patterns influence how organizations balance accessibility with safety.
The takeaway: enabling broad access to AI agents can unlock practical benefits, but it must be paired with robust controls to manage the attendant risks.
