The Executive Yuan yesterday passed the Guidelines for the Executive Yuan and Affiliated Agencies on Using Generative Artificial Intelligence (行政院及所屬機關(構)使用生成式AI參考指引) and banned the use of artificial intelligence (AI) programs in drafting of classified documents.
The guidelines come into effect from today.
The Executive Yuan said it developed the guidelines to ensure that the use of generative AI programs would not harm national security, or information security, human rights, the rights to privacy, morality and ethics, and would not violate the law.
Agents or personnel should make the final judgement on whether information generated by generative AI programs should be used, and the programs should not replace or supersede the personnel’s autonomous thought, creativity or social interaction, the guidelines say.
All classified documents — defined by the Executive Yuan’s Handbook on Handling Documents (文書處理手冊) — must be prepared by the agent or personnel responsible and should not be drafted by generative AI programs, they say.
Personnel or agents should not provide generative AI with information that should be kept secret such as personal information or information that has not been authorized to be made public, nor should they input queries into generative AI programs for possibly classified information or personal information, they say.
However, personnel in agencies with closed network systems and generative AI may, after confirming that the system environment is secure, use generative AI for queries, based on levels of security clearance, they say.
Government agencies should not fully trust information produced by generative AI, nor should they base government policy or administrative decisions on unverified information provided by generative AI, the guidelines say.
There should be due levels of transparency when government agencies use information produced by generative AI, they say.
The use of AI should follow regulations on information safety, protection of personal information and copyright laws, they say.
Individuals must be aware that the data provided by generative AI could infringe on copyright laws or personal rights, and all government agencies should draft and implement internal regulations regarding the use of generative AI programs, the guidelines say.
According to the guidelines, legal personae and groups of individuals obtaining government project bids should observe these guidelines and the internal regulations of the agency issuing the project.
The guidelines would be applicable to the use of generative AI at state-owned businesses, state-funded schools, juridical persons and state-funded non-departmental public bodies.
National Science and Technology Council Deputy Minister Lin Minn-tsong (林敏聰) said there is no time line to introduce an essential AI act, and drafting such an act depends on AI development domestically and abroad.
