《TAIPEI TIMES》 Science council working on AI guide: minister

SCRUTINIZING CHATGPT: Minister of Digital Affairs Audrey Tang announced the move after some local governments touted plans to integrate the application

By Shelley Shan / Staff reporter

請繼續往下閱讀...

The National Science and Technology Council （NSTC） is compiling guidelines on the use of artificial intelligence （AI） in government agencies after several local governments announced plans to use ChatGPT to offer services to residents, Minister of Digital Affairs Audrey Tang （唐鳳） said yesterday.

Tang made the remark at a meeting of the legislature’s Transportation Committee as she and other officials briefed lawmakers on measures that the government would take to prevent personal data breaches as people increasingly file their taxes online and conduct business using ChatGPT.

Taoyuan is planning to use ChatGPT in its 1999 city government hotline, while Tainan has developed an auxiliary teaching module using the AI app.

Democratic Progressive Party （DPP） Legislator Lee Kun-tse （李昆澤） said many governments and private corporations around the world have begun using ChatGPT, but countries such as Italy have banned the use of the technology due to privacy concerns.

He asked Tang whether Taiwan would ban civil servants from accessing ChatGPT or other generative AI applications to conduct government affairs.

Tang said Italy lifted the ban on ChatGPT after the developer, OpenAI, implemented changes Rome asked for.

The NSTC is compiling a guidebook on the use of generative AI applications in government agencies, Tang said.

“Two major principles will be stated in the guidebook. One is that government workers must not submit content generated by generative AI systems without scrutinizing it. Different from content found through search engines, content created by generative AI systems can come out of nowhere. They could be untruthful statements that appear to be logical,” she said.

Second, personal data breaches can be prevented if government departments only use AI systems when they are not connected to the Internet and instead use local servers over which they have full control, she said.

Regarding the risks of data leaks at companies using AI, Tang said that corporations should always be careful not to leak confidential information to the public.

“The problem is that most corporations depend too much on AI language models available online because they are convenient to use. If corporations could themselves develop similar AI applications that are just as convenient to use as ChatGPT, they need not access the online chatbot and risk having data leaks,” she said.

The National Institute of Cyber Security is building a Taiwanese AI framework and is planning to work with the Industrial Technology Research Institute （工業技術研究院） on related projects in the second half of this year, she said.

The NSTC is scheduled to introduce a “trustworthy AI dialogue engine” to the public by the end of this year, Tang said, adding that the National Institute of Cyber Security would be in charge of authenticating the intermediary versions of the dialogue engine.

“Any organization should use the local technology or work with trusted local cloud providers so as not to upload all data to the free public cloud, which will lead to concerns about data leaks,” she said.

新聞來源：TAIPEI TIMES