Regarding the legislative progress of the draft Artificial Intelligence Basic Act, Yi-Jing Lin, Deputy Minister of the Ministry of Digital Affairs (MODA), stated that the draft has been submitted to the Executive Yuan. However, due to recent challenges like the typhoon disaster and trade negotiations, there has been a lack of time to process it, and he hopes to send the draft to the Legislative Yuan for review as soon as possible. Regarding the regulatory mechanism, which is currently the biggest point of disagreement between the ruling and opposition parties, MODA plans to follow the U.S. model by establishing a risk classification framework within subsidiary legislation.
The joint committee meeting of the Legislative Yuan's Education and Culture and Transportation Committees completed the preliminary review of the draft Artificial Intelligence Basic Act on August 4. A total of 19 versions were proposed by lawmakers from both the ruling and opposition parties. Due to divergent opinions, only three articles—mandating the government to promote international cooperation, jointly promote innovative applications with the private sector, and establish data openness mechanisms—were passed, while all other key articles were reserved for cross-caucus negotiation in the full session.
An overview of the current legislative disagreements shows the main differences lie in the regulatory mechanism and whether to stipulate penalties. For example, Kuomintang (KMT) legislators including Chiung-Ying Yang and Tsung-Hsien Wu proposed that the government should establish principles for AI risk classification and determination, and implement risk assessment and regulatory mechanisms for appropriate control. KMT legislator Yu-Chen Hsu advocated for establishing a risk assessment mechanism based on low, medium, and high risks, and implementing supervisory measures such as random inspections and periodic safety tests.
KMT legislator Shih-Bao Lai proposed that the government should establish a safety assessment and certification system to ensure the security and reliability of AI systems. KMT legislators Yi-Fong Hsieh and Chien-Chi Lin suggested setting up a special audit mechanism, the "AI Special Committee," to handle ethical compliance concerns.
Furthermore, bipartisan legislators all advocated for promoting an internationally aligned risk grading or risk classification framework. They proposed mechanisms such as standards, verification, testing, labeling, disclosure, traceability, or accountability to enhance application trustworthiness.
As the Executive Yuan's version of the Artificial Intelligence Basic Act draft has not yet been formally released, there is widespread attention on the executive branch's planned direction. Deputy Minister Yi-Jing Lin told the Central News Agency that Europe and the United States have different concepts regarding AI risk management. For instance, EU law explicitly prohibits harmful practices such as using AI systems for biometric classification, whereas the U.S. believes excessive regulation should be avoided to prevent hindering innovation. Therefore, the biggest difference between the draft submitted by MODA and the NSTC's previous draft notification is the change from "risk grading" to "risk classification."
SSL 256bit transmission encryption mechanism