关键词: 教育智能体, 生成式人工智能, 人工智能风险, 人工智能伦理

Abstract: With the explosive iteration of generative artificial intelligence technology, the application of agents in education is becoming increasingly widespread, and the resulting ethical issues deserve attention. From the perspectives of agent application and regulation, risks in educational scenarios-spanning the teaching process, learning assessment, data content, and regulatory governance-stem primarily from the trend toward technological democratization, shortcomings in the regulatory system, and insufficient literacy among teachers and students. To formulate scientific risk-mitigation strategies, it is necessary to systematically review and compare principles, norms, and guidelines issued by international organizations, national governments, and universities concerning the application of AI in education. Strategies should be structured from four pillars: technical governance, institutional development, capacity building, and value adherence. Concrete measures include: building trustworthy, transparent, and controllable systems; improving access, grading, and filing systems; enhancing the AI collaboration capability of teachers and students; and ensuring that educational agents remain human-centered.

Key words: educational agent, generative artificial intelligence, AI risk, AI ethics