If You Use Generative AI, You Must Disclose the Model and Version... Researchers Responsible for National R&D

[Financial News, Jeju = Jeong Yong-bok, Reporter] New research ethics standards require researchers to disclose in their final results which generative artificial intelligence (AI) model and version they used, and where they used it, during national R&D projects involving tools such as ChatGPT. In Jeju, determining how to incorporate generative AI standards into existing research ethics and security systems at universities and government-funded research institutes conducting national R&D has emerged as a new challenge for research management.
The Ministry of Science and ICT and other sources said on the 28th that the Steering Committee of the Presidential Advisory Council on Science and Technology (PACST) recently deliberated on and finalized the AI Research Ethics Guidelines for National R&D.
The basic principle presented by the government is clear. AI's role in national R&D is limited to assisting researchers as a tool, while the responsibility and rights associated with the final results belong to the researchers who used AI.
The seven standards researchers must follow are fact-checking, independent judgment, transparency, management of result reliability, compliance with policies and regulations, security, and protection of personal information.
Researchers may not use papers, statistics or references provided by AI without verification. They must personally confirm that the materials actually exist and that their contents are accurate. Researchers must also review the logical validity of AI-generated analyses and code before using them, and apply their own professional judgment.
The biggest change is the disclosure of how AI was used. If researchers use AI during the research process, they are advised to disclose the model and version used, along with the purpose and scope of its use, when submitting or presenting their final research results. Examples of information subject to disclosure include not only information searches and literature reviews, but also data inference and organization, translation, and grammar and typo checks.
The AI use disclosure form requires researchers to provide the tool used, its model and version, the details of its use, and the period of use. Generative AI tools such as ChatGPT and Claude were cited as examples.
The government has not recognized AI as an author or co-researcher. Because AI cannot bear legal or moral responsibility, it cannot be an author of a paper or an inventor named in a patent. The researcher who used AI must remain responsible for the research idea, analysis and conclusions.

Research security standards are stricter. Researchers are instructed not to indiscriminately enter technologies pending patent applications, unpublished research materials or confidential data provided by companies into external commercial AI services. If research materials containing personal information must be used with AI, necessary measures such as de-identification must be taken first.
In national R&D project evaluations and paper reviews, evaluators, as well as researchers, are subject to separate duties of caution. Evaluators must not enter research proposals, final reports or papers under review into external AI services. Such materials could contain unpublished research ideas, technical information and personal data that might be transferred to external servers.
The standards also cover a new form of misconduct aimed at manipulating evaluation results with AI. A representative example is a hidden prompt, in which researchers insert text or commands that are difficult for people to notice into a research proposal or paper to induce AI to produce an evaluation favorable to them. The government specified that such acts, which seek to evade legitimate evaluation criteria or distort review results, may constitute research misconduct and may be subject to sanctions under relevant laws and regulations.
The new guidelines also intersect with existing research ethics systems in a number of ways at research sites in Jeju.
The regulations published by the Industry-Academic Cooperation Foundation of Jeju National University include guidelines for research notes under national R&D programs, security management guidelines for R&D programs overseen by the Ministry of Science and ICT, the university's own research ethics regulations, and security management guidelines for research projects and results.
Because the new guidelines call on each R&D institution to establish relevant regulations, guidelines and security environments, efforts to incorporate AI use standards into existing research ethics and security rules are likely to follow.
It is also noteworthy that government-funded research institutes in Jeju conduct a significant number of demonstration projects involving research data and companies' technical information.
The Jeju Global Research Center has been conducting a project titled "Development of Core Technologies for Expanding the Capacity to Accommodate Variable Renewable Energy through Multi-Sector Coupling," with the participation of 22 organizations, including government-funded research institutes, universities, Jeju Island, Jeju Energy Corporation (JEC) and private companies. The six-year project has a budget of 46.1 billion won and is a representative example of large-scale energy demonstration research being conducted in Jeju.
The Jeju Technology Commercialization Division of the Korea Institute of Industrial Technology also focuses on research areas including the transition to clean energy, digital manufacturing and future mobility demonstrations.

When generative AI is used in such national R&D projects to assist with literature reviews, coding or experimental data analysis, the new guidelines' principles of disclosure, verification and security apply. In particular, research security management may become a practical issue when researchers conduct joint projects with companies and must avoid entering confidential materials received from them or technologies awaiting patent applications into external AI services.
Jeju Technopark is also expanding research and commercialization projects that combine AI and biotechnology. An AI-Bio Convergence Strategy Seminar held this year discussed a database of biological materials, AI services, autonomous laboratories, personalized services using lava seawater and health examination data, and language models specialized in biotechnology.
Not all Jeju Technopark projects are directly covered by the new national R&D guidelines. Whether they are covered depends on whether each individual project qualifies as national R&D. However, as more projects combine health information, research data and corporate materials with AI, the principles of protecting personal information and data, and assigning responsibility for verifying AI-generated results, will inevitably become important management standards for regional research and industry as well.
Researchers' use of AI is already rising rapidly. A Wiley survey of about 2,400 researchers worldwide last year found that the proportion saying they used AI at work increased from 57% in 2024 to 84% in 2025. The share using AI for research and publishing also rose from 45% to 62%.
In an Elsevier survey of 3,000 researchers, the proportion saying they used AI at work rose from 37% in 2024 to 58% last year. By contrast, only 32% of researchers said their institutions had well-established AI governance, while 27% said they had received sufficient training in AI use.
These figures provide the backdrop for the new guidelines, which seek to have researchers disclose their use of AI, verify its results themselves and protect research materials rather than prohibit AI use.
The standards serve as guidelines for researchers to use AI responsibly in national R&D settings. The Ministry of Science and ICT plans to distribute the guidelines and explanatory materials to R&D institutions and continue improving them by reflecting feedback from the field.
Hong Sun-jeong, director general of the Research Performance Evaluation Policy Bureau at the Ministry of Science and ICT, said, "AI is a useful tool that can dramatically improve research efficiency, but final responsibility for it always lies with the researcher. I hope this will serve as an opportunity to establish a culture of responsible AI use and ensure the integrity of national R&D."


[email protected] Jeong Yong-bok Reporter