Free China Movement (FCM) uses modern research technologies, including artificial intelligence, to help examine large, multilingual, and historically dispersed bodies of information.
AI can dramatically increase the scale of research. It does not replace evidence, editorial judgment, or human accountability.
This policy explains the principles governing its use.
Why FCM Uses AI
Research concerning China, democracy movements, political prisoners, constitutional development, human rights, and historical events can involve thousands of documents distributed across decades, countries, archives, languages, and media formats.
AI-assisted tools may help researchers discover potentially relevant materials; organize large collections; compare documents; identify names, dates, organizations, and relationships; search multilingual sources; assist translation; identify apparent inconsistencies; construct preliminary chronologies; classify archival materials; summarize material for further examination; and assist drafting and editing.
AI Output Is Not Evidence
An AI system’s statement is not a primary or secondary historical source merely because the system states it confidently. AI systems can misunderstand sources, merge separate people or events, mistranslate terminology, misstate dates, omit important context, reproduce errors from secondary sources, and generate citations or factual assertions that do not exist.
For this reason, consequential factual claims should be traced to identifiable underlying evidence whenever reasonably possible.
Human Responsibility
AI may assist the work. It does not assume responsibility for it. Material published under FCM’s institutional name remains subject to human responsibility for editorial judgment, sourcing, corrections, and publication decisions. FCM does not treat the use of an AI system as a substitute for appropriate review.
Human Review Should Be Risk-Based
Greater scrutiny should be applied to claims involving accusations against identifiable individuals; political-prisoner or detention status; deaths or casualties; criminal allegations; disputed historical events; quotations; sensitive personal information; consequential translations; legal or constitutional interpretation; and conclusions that could materially affect a person’s reputation or safety.
Source Verification
Where an AI-assisted process identifies an important fact, quotation, document, or citation, researchers should seek the underlying source rather than relying solely on the AI-generated description. When the original evidence cannot be located, the claim should be appropriately qualified, omitted, or identified as unverified rather than presented as established fact.
AI-Assisted Translation
AI may assist multilingual publication and research. Machine-generated translation should receive review proportionate to its significance, particularly for quotations, legal terminology, political terminology, historical documents, names, and statements whose meaning could materially change through translation.
AI-Assisted Drafting
AI tools may assist with outlining, drafting, editing, comparison, or formatting. The presence of AI assistance does not reduce the standards applied to factual claims. FCM evaluates published material based on its evidence, accuracy, transparency, and editorial responsibility—not on whether a particular sentence was initially typed by a person or generated with technological assistance.
Historical Research and AI
AI is particularly useful for discovering connections across large archives, but historical inference requires caution. A suggested relationship between people, organizations, documents, or events should not be represented as historical fact until the underlying evidence supports it. Conflicting accounts should not be automatically reconciled merely because an AI system proposes a convenient narrative.
Transparency
FCM maintains this public policy so readers can understand that AI-assisted methods may form part of our research and publishing workflow. Where AI plays an unusually substantial or consequential role in a particular research project, additional methodological disclosure may be provided on that project.
Corrections
AI-assisted work is covered by the same Corrections Policy as other FCM publications. If an AI-assisted process contributes to a factual error, the fact that technology was involved does not excuse the error or reduce FCM’s responsibility to correct it.
Privacy, Security and Sensitive Sources
AI tools should be used cautiously with confidential, unpublished, personally identifying, or security-sensitive information. Information that could expose dissidents, political prisoners, witnesses, family members, confidential sources, or other vulnerable people to foreseeable harm should not be unnecessarily submitted to external AI systems. Protection of vulnerable sources takes precedence over research convenience.
The Principle
Use AI to expand what researchers can discover—not to lower the standard of what FCM claims as fact.
Technology can accelerate discovery. Evidence establishes the record. Human judgment remains accountable for publication.
Related policies: Editorial Standards · Research Methodology · Corrections Policy
Last reviewed: September 2026
Free China 