China’s Top Academy Bans 10 Types of AI-Assisted Research
In a landmark move to regulate artificial intelligence in academia, the Chinese Academy of Social Sciences (CASS) has issued a comprehensive 37-article regulation prohibiting 10 categories of AI-assisted research activities, marking what analysts believe is the first institutional framework of its kind from a major national social science academy worldwide.
The “Basic Norms for AI-Assisted Research at the Chinese Academy of Social Sciences (Trial)” — designated as document number 社科研字〔2026〕6号 — was released on July 9 and establishes clear boundaries for AI use across CASS’s network of more than 30 research institutes covering economics, history, philosophy, law, and international studies. The rules, reported by People’s Daily, aim to balance technological adoption with academic integrity, data security, and ideological compliance.
A Delicate Balance: Permissive Yet Restrictive
The regulation is notable for its dual nature. While it enumerates 11 permitted use cases — including topic selection, research design, knowledge sorting, surveys, archaeological work, and international academic exchange — it simultaneously draws 10 rigid “red lines” that researchers must not cross.
According to the full text published by CASS, the framework rests on four core principles: human-led, AI-assisted (人主智辅); quality and efficiency improvement; truthful disclosure; and safety compliance.
Tang Xiaofeng (唐晓峰), head of the CASS Research Bureau, explained the philosophy behind the regulations in an in-depth interview with China Social Sciences News. “Setting rules requires both opening paths and drawing red lines while holding the bottom line,” Tang said. “The 10 red lines set up defenses from macro-level political bottom lines to micro-level research operations, delineating behavioral禁区 for all researchers.”
The 10 Prohibited Categories
The banned activities span four risk dimensions, addressing concerns from national security to academic fraud:
Political and ideological risks top the list, with a prohibition on using AI for activities that violate socialist core values, endanger national security, or breach laws and academic ethics. Researchers are also barred from generating or disseminating content that distorts facts or misleads the public.
Data security is a major focus. The regulations forbid inputting classified information into AI systems, illegally scraping data, or leaking citizens’ personal information — a reflection of growing concerns about state secrets and data sovereignty in the AI era.
Academic integrity violations receive the most extensive treatment. The rules prohibit using AI to forge or tamper with data, generate false cases, or fabricate evidence. A key innovation is the requirement for “truthful disclosure” — researchers must retain prompts, interaction logs, and input-output records to create a verifiable audit trail. Deliberately concealing AI usage is defined as a new type of academic fraud.
The regulations also ban AI from completely replacing research work, using AI for plagiarism or content laundering, mass-producing templated peer review opinions, and churning out low-quality repetitive research outputs that damage the academic ecosystem.
A Gradual, Systematic Approach
The CASS regulations did not emerge overnight. They build upon earlier efforts including the CASS History Research Institute’s generative AI guidelines issued in February 2025, and subsequent policies from seven CASS sub-units. The State Council’s “AI+” Action Opinion in March 2025 and the “AI + Education” Action Plan in 2026 provided broader policy impetus.
Tang emphasized that the drafting process spanned six months and included comprehensive research, cross-disciplinary drafting teams, and multiple rounds of consultation. “Researchers are the主体 of academic思考,” he said. “The core output of philosophical and social science research is original thought, independent viewpoints, and systematic theory — these cannot be replaced by AI.”
Implications for Global Academic AI Governance
The CASS framework is believed to be the first comprehensive institutional AI regulation from a major national social science academy anywhere in the world. As China’s premier think tank, CASS’s approach is likely to influence policies at other Chinese universities and research institutes.
The regulations arrive amid a broader global reckoning with AI’s impact on research integrity. Universities worldwide have been updating their AI usage policies, the European Union’s AI Act established a comprehensive regulatory framework in 2024, and academic journals across disciplines continue to refine their stances on AI-generated content.
What to Watch For
Several questions remain unanswered. Enforcement mechanisms and specific penalties for violations have yet to be detailed. The “trial” designation suggests the framework will evolve as AI technology advances. Most significantly, observers will be watching whether other Chinese research institutions — and potentially institutions in other countries — adopt similar frameworks.
Tang acknowledged the experimental nature of the regulations. “If only restrictive norms are issued without配套 support guarantees, researchers may develop the misconception that they are ‘only restricted, not supported,’” he said. The document includes six supporting measures covering infrastructure, evaluation reform, interdisciplinary development, data resources, cross-sector collaboration, and talent cultivation.
As AI continues to reshape academic research worldwide, CASS’s bold regulatory experiment offers a potential model — and a cautionary tale — for institutions grappling with how to harness AI’s power without sacrificing the human intellect at the heart of scholarly inquiry.