As generative artificial intelligence accelerates its way into classrooms and laboratories, a quiet yet profound transformation is unfolding across universities worldwide. Students are turning to AI to support their learning and spark new ideas, even as they confront concerns about “mental inertia” and the outsourcing of cognitive effort. Faculty members, meanwhile, are exploring new avenues for AI-empowered instruction while grappling with questions of boundaries and ethics.
Recently, Tsinghua University released the Tsinghua University Guiding Principles for the Application of Artificial Intelligence in Education (“the Guiding Principles”), the institution’s first comprehensive, university-wide framework that sets systematic, multi-level guidance and norms for the use of AI across campus.
Five Core Principles: Defining the Red Lines—and the Pathways Forward—for AI in Education
The Guiding Principles are structured into three parts—“General Provisions,” “Teaching and Learning,” and, “Theses, Dissertations and Practical Achievements.” Together, they offer a top-level framework supported by detailed, scenario-based guidelines that cover the core contexts of teaching and academic research.
The “General Provisions” outline Tsinghua’s stance of being “proactive yet prudent” in its approach to AI and establish five core principles: principal responsibility, compliance and integrity, data security, prudence and critical thinking, and fairness and inclusiveness. These principles affirm that AI must remain an auxiliary tool and that teachers and students continue to be the primary agents in teaching and learning. The framework calls for proper disclosure of AI use and prohibits academic misconduct; forbids the use of sensitive, classified, or unauthorized data when training or operating AI models; urges vigilance toward AI “hallucinations” and stresses multi-source verification to guard against the cognitive complacency that can result from overreliance on AI; and underscores the importance of identifying and mitigating algorithmic bias and the digital divide to ensure that technology can be used for public good.
The “Teaching and Learning” section advises instructors to determine how AI should be used according to the goals of each course, clearly explain the Guiding Principles to students at the start of the semester, and remain responsible for the AI-generated teaching materials. The section also encourages teachers to guide students toward a critical and well-rounded understanding of AI, helping them develop essential competencies. Students, in turn, are encouraged to explore AI as auxiliary aids within the boundaries set by each course, but they are strictly prohibited from copying or mechanically paraphrasing AI-generated text, code, or other output as academic submissions.
For graduate students, the section on “Theses, Dissertations and Practical Achievements” highlights that AI may not replace the academic training and intellectual labor that students are expected to complete independently. The use of AI for ghostwriting, plagiarism, fabrication, or other forms of misconduct is strictly forbidden. Supervisors are required to provide clear guidance on AI use and maintain full-process oversight to ensure both the integrity of academic training and the originality of students’ work.
According to Wang Shuaiguo, Director of the Online Education Center and primary drafter of the document, the Guiding Principles also reserve ample space for future development. Beyond current teaching and research scenarios, the framework is designed to support future extensions into academic research, administrative services, and other emerging applications. “We hope this will not become a document that restricts innovation,” he said. “It should be a living system—one that continues to grow as the technology evolves.”
Two Years in the Making: Responding to Frontline Needs in Teaching and Learning
The development of the Guiding Principles reflects Tsinghua University’s close observation of AI’s growing presence in education, its extensive research, and the gradual formation of shared consensus.
“Technology often matures earlier than the risks it brings become visible,” noted Professor Li Manli of the School of Education. She emphasized that universities need time to fully explore and discern the opportunities and challenges that AI brings to real teaching and learning contexts.
Since 2023, a wave of innovations has taken shape across Tsinghua’s campus -- including AI-empowered courses, disciplinary knowledge engines, agent instructors, and the all-in-one campus companion Qing Xiaoda. These developments have also brought new questions: whether AI-assisted assignments should be treated as academic misconduct, how responsibility should be defined when AI is used to grade coursework, and where the boundaries lie when graduate students rely on AI to process data or generate code.
“These real issues had to be addressed.” Work began immediately. A research team led by Professor Li conducted a global survey of AI education guidelines, systematically analyzing 70 documents released by 25 universities across the United States, Europe, and the Asia-Pacific region.
In the summer of 2024, the University launched the drafting of the Guiding Principles, led jointly by the Undergraduate Academic Affairs Office, the Graduate School, the School of Education, and several departments and academic units.
At the same time, the University officially initiated a broad internal consultation process. From March to May 2025, the project team interviewed more than one hundred students and frontline instructors from disciplines spanning the humanities, sciences, engineering, and medicine. Many expressed the hope that the University would clarify the boundaries of academic integrity and offer clear guidance on AI usage.
“Different disciplines and different habits of AI use naturally led to different viewpoints. Every discussion was intense,” recalled Wang. “But it was through these exchanges that we gradually reached consensus.”
Feedback from teachers and students was directly woven into the Guiding Principles. Multiple rounds of review by the Teaching Committee and the Degree Evaluation Committee further ensured the document’s scientific rigor and practical feasibility.
“Through parallel tracks of pilot experimentation and research-driven inquiry, Tsinghua has taken a measured approach to the new challenges of the AI era.” In Li’s view, the Guiding Principles represent a pragmatic effort to find the broadest common ground between “active embrace” and “prudent use.”
The Future Is Here: Building a Sustainable Ecosystem for AI in Education
The release of the Guiding Principles marks a new stage in Tsinghua University’s exploration of AI–education integration—one that moves from “technology advancement” to “institutional support.” This shift is rooted in the University’s sustained and systematic groundwork in AI-enabled teaching over recent years.
As early as 2023, Tsinghua launched its first pilot programs for AI-empowered courses. After more than two years of development, over 390 courses now incorporate AI deeply into teaching practice, spanning ten functional areas that include AI learning companions, AI teaching companions, and AI assistants for lesson preparation. These pilots have generated a rich pool of practical experience.
The three-layer decoupled architecture
A key technical backbone in this process has been the University’s self-developed three-layer decoupled architecture. By flexibly connecting multiple mainstream AI models at the foundational layer, building precise discipline-specific knowledge engines in the middle layer, and supporting diverse teaching applications at the top layer, the architecture reduces dependence on any single model and provides a scalable solution for integrating AI into instruction in a scientific and efficient manner.
At the same time, Tsinghua has built a multi-level AI general education ecosystem encompassing general education courses, minor programs, and certificate pathways to enhance students’ AI literacy systematically. The University’s all-in-one learning companion, Qing Xiaoda, functions as an always-available growth assistant, offering personalized support throughout the entire learning journey.
Importantly, the Guiding Principles are designed not to constrain innovation; rather, they aim to enable it. The General Provisions explicitly state that the University encourages and supports faculty and students to pursue innovative applications of AI in teaching and learning and will recognize and promote exemplary practices. In this sense, the Guiding Principles not only establish “red lines” but also illuminate “green lights” for responsible, meaningful experimentation.
Looking ahead, Tsinghua will continue to advance the implementation and evolution of the Guiding Principles through the AI literacy platform, teaching workshops, and interdisciplinary discussions.
“What we confront is not merely the question of how to employ a particular tool, but rather a fundamental redefinition of the very nature of education in the age of artificial intelligence.” With the Guiding Principles now in place, Tsinghua aims to guide its community toward responsible, informed AI use to foster a positive, healthy, and sustainable ecosystem for educational innovation—preparing the next generation of talents to harness intelligent technologies and innovate alongside them.
Editors: Gillian, Ma Mingwei, Li Han, John Paul Grima
