Newsletter
NSTC announced the draft Artificial Intelligence Basic Act
In view of the rapid development of artificial intelligence (“AI”) in recent years, and in order to establish a legal environment conducive to the development of AI technology and applications while taking into account human rights and all relevant risks, the National Science and Technology Council (“NSTC”) announced the draft Artificial Intelligence Basic Act (“Draft Act”) on July 15, 2024. Below is a summary of the Draft Act:
1. To differentiate AI from other software systems, Article 2 of the Draft Act defines AI as “a machine-based system that can operate autonomously and, through input or sensation, realize the output that would affect the physical or virtual environment (such as predictions, content, recommendations, or decisions) for explicit or implied objectives by means of machine learning and algorithms.”
2. To mitigate the risks incurred from AI innovation and development while taking into account social welfare and digital equality, Article 3 of the Draft Act stipulates that the government shall adhere to the following fundamental principles when promoting the R&D and application of AI:
(1) Sustainable Development and Welfare: Consider social fairness and environmental sustainability, provide appropriate education and training, and reduce potential digital divide.
(2) Human Autonomy: Uphold human autonomy, respect fundamental rights and cultural values, and implement basic human-centric values.
(3) Privacy Protection and Data Governance: Properly protect the privacy of personal data and promote the release and reuse of non-sensitive data.
(4) Information Security and Safety: Establish information security measures to ensure system robustness and safety.
(5) Transparency and Explainability: AI-generated output should be properly disclosed or labeled to facilitate the assessment of potential risks and to understand the impact on stakeholders’ rights and benefits.
(6) Fairness and Non-Discrimination: Avoid algorithm bias and discrimination as much as possible.
(7) Accountability: Ensure appropriate accountability, including internal governance responsibilities and external social responsibilities.
3. Articles 4 through 17 of the Draft Act further set out the following policy objectives based on the aforementioned fundamental principles and the government’s priorities:
(1) The government shall adequately plan the overall allocation of resources and handle matters about subsidies, commissions, investments, incentives, assistance, and guidance for AI-related industries or provide financial incentives such as taxes or financing.
(2) The government shall endeavor to improve the adjustment and adaption of laws and regulations to ensure that their interpretation and application comply with the aforementioned fundamental principles and do not hinder the provision of new technologies and services.
(3) The central competent authorities in charge of the relevant industries may establish or complete existing innovative experimental environments (i.e., regulatory sandboxes) for AI innovative products or services.
(4) The government should cooperate with the private sector to promote AI innovation and application through public private partnerships.
(5) The government shall endeavor to promote AI-related international cooperation and participate in international joint development and research programs.
(6) The government shall continue to promote AI education at all school levels, within industries, and across society, government agencies, and public institutions.
(7) The government shall prevent AI applications from causing personal or property damage, destruction of social order or eco-environment, conflict of interest, bias, discrimination, false advertising, misinformation, falsification, or other problems that violate relevant regulations. The Ministry of Digital Affairs (“MODA”) and other relevant authorities may provide or recommend assessment and verification tools for the central competent authorities in charge of the relevant industries to deal with the aforementioned matters.
(8) The MODA shall promote a risk classification framework for AI that aligns with international standards so that the central competent authorities in charge of the relevant industries may stipulate their risk classification regulations for those industries under their supervision.
(9) The government shall, based on AI risk classification, assess potential vulnerabilities and abuses through standards, regulations, or guidelines to enhance the verifiability and human control of AI decision-making.
(10) Based on AI risk classification, the government shall, through standards, verification, inspection/testing, labeling, disclosure, traceability, accountability, or other mechanisms, establish regulations concerning the conditions, responsibilities, remedies, compensation, and insurances for AI applications, specifying attribution of responsibilities and criteria of attribution. However, any activities prior to applications shall be exempt from the aforementioned accountability regulations to facilitate AI technology innovation and development.
(11) The government shall prevent skill gaps, protect labor rights, and provide employment measures for those becoming unemployed due to the use of AI based on their work capabilities.
(12) The competent authority for personal data protection (i.e., the future Personal Data Protection Commission) shall assist the central competent authorities in charge of the relevant industries to promote “data protection by default and by design” measures or mechanisms (e.g., the Guidelines on Privacy Enhancing Technologies issued by the MODA).
(13) The government shall establish a mechanism for open data, data sharing, and data reuse to enhance the availability of data used by AI and regularly review and adjust/adapt relevant laws and regulations.
(14) The government shall endeavor to improve the quality and quantity of data used by AI and ensure that the training results uphold national multicultural values and respect intellectual property rights.
(15) To comply with AI fundamental principles, when using AI for its operations or the provision of services, the government shall conduct risk assessments and plan risk response measures; government agencies and public institutions shall formulate rules of use or an internal control and management mechanism according to the nature of the operations involving the use of AI.
(16) After the implementation of the Draft Act, the government authorities shall review and adjust their job duties, operations, and laws/regulations in accordance with the provisions of the Draft Act to achieve its objectives; before the enactment or amendment of the relevant laws/regulations, in the absence of existing regulations, the central competent authorities in charge of the relevant industries shall, in collaboration with the NSTC, interpret and apply the provisions of the Draft Act.
Similar to other “basic acts” (such as the Road Traffic Safety Basic Act and the Science and Technology Basic Act), the Draft Act aims to stipulate the government’s tasks to promote AI development and does not impose any specific regulatory requirements for AI. However, it is worth observing the developments concerning the Draft Act and how the central competent authorities in charge of the relevant industries will adjust/adapt their laws/regulations and establish their risk classification regulations after the enactment of the Draft Act.