China Academy unveils top 10 trends for AI agents in 2026

China's Academy of Information and Communications Technology released its top 10 trends for AI agents in 2026, offering a forward look at tech breakthroughs, real-world use, and ecosystem building. The list covers agent infrastructure, collaboration, engineering, learning, memory, skills, product innovation, payment protocols, trustworthiness, and full-stack evaluation.

Agent infrastructure is seen as a "silicon incubator" that supports development and deployment, pushing for standardization and cloud-edge teamwork. Agent collaboration aims to connect different systems and agents, bridging virtual and physical worlds. Agent engineering focuses on the full lifecycle, creating a closed loop that's repeatable, observable, and optimizable.

Learning and evolution let agents improve through ongoing interaction, shifting from passive commands to active self-growth. Agent memory acts as an "experience warehouse" for context and long-term adaptation. Skills package tasks into reusable units for better accuracy. New products are moving beyond simple Q&A to cover office, production, and daily life.

Payment protocols standardize autonomous transactions. Trustworthiness relies on data, tech, and evaluation to ensure reliability. Full-stack evaluation measures tech ability, value, and efficiency, with the Academy's benchmarks continuing to evolve.