Mon, Sep 16, 2024 page2

• Draft AI act lacking key requirements, group says

  STILL EARLY: The basic act should have some flexibility, focusing on general principles, while government agencies can still place rules on the technology, a law professor said

  • By Shelley Shan / Staff reporter
  • • 
    • 
    • 

  The National Science and Technology Council’s (NSTC) draft artificial intelligence (AI) basic act (人工智慧基本法) lacks the required procedures that AI researchers and developers should follow and could contradict existing regulations on personal data protection, financial supervision and medical equipment, the Taiwan Association for Human Rights said.

  After the Ministry of Digital Affairs in 2019 published guidelines governing the research and development of AI, the council this year announced the draft AI basic act.

  The draft was released amid growing calls from the international community for governments to address the risks AI development pose and to safeguard basic human rights.

  

  A figure is positioned above a keyboard and a display showing the words “artificial intelligence” in an illustration photograph taken on May 4 last year.

  Photo: Reuters

  While a large part of the draft act focuses on the role of government in facilitating the development of AI, the procedures that government agencies should follow when developing and applying AI remain unclear, the association said.

  A few problems have emerged from the government’s use of AI, such as when the Judicial Yuan used AI to draft verdicts of criminal cases, while the government has yet to disclose the results of the Ministry of Justice’s use of AI to evaluate prisoners’ risks of recidivism, it said.

  The National Health Insurance Administration and Google this year jointly developed an AI system to evaluate the risks of diabetes, but documents related to the partnership did not show the sources of the data used to train the system, it said.

  Additionally, a draft act governing the management of health and social welfare data (衛福資料管理條例) has yet to be passed by the legislature, the association said.

  Unlike the EU, which regulates AI based on risk levels involved in its use, Taiwan’s draft AI act only requires the digital ministry to stipulate an AI risk framework without clearly defining relations between the council and other agencies, it said.

  The draft act also does not include regulations on basic human rights protections and remedial measures, it added.

  Whether the act can protect personal privacy and data security and avoid the discrimination of specific groups and individuals are questionable, it said.

  Although the draft lists seven major principles, it does not have corresponding rules of implementation, the association said.

  Regarding the principles of privacy protection and data management, the draft allows the reuse of nonsensitive data, such as data unrelated to medical history, genetic information, sexuality, health examinations and criminal records, which could contradict the Personal Data Protection Act (個人資料保護法), it said.

  “We strongly advise the government to first clarify the relation between the draft act and other regulations, and revise the legal framework to ensure that all measures designed for safety and to protect human rights could be thoroughly implemented. The public and private sectors should have rules to abide by in the development and application of AI,” the association said.

  In response, the NSTC said the draft act does not allow AI developers and researchers to disobey personal data protection, financial supervision and medical equipment laws.

  National Yang Ming Chiao Tung University law professor Carol Lin (林志潔) said that Taiwan, much like the US, is in the research-and-development phase of AI.

  Having some flexibility in the basic act is not completely unreasonable, she said.

  “The focus of a basic act is to offer general principles, guidelines, directions and flexibility as an industry grows,” she said. “Different government agencies can still stipulate rules to counter the effects of AI on industries under their supervision. The draft act would not contradict other laws.”

  The government should set up an experimental arena with controllable risks for start-ups to test their AI-related services, products and business models, Lin said, adding that this would help the government revise and stipulate AI laws.

  The government used the “financial regulatory sandbox” model to regulate financial technology at the initial phase, she said.
• Most Popular

  • 1

    Bukele proposes prisoner swap with Maduro

  • 2

    HK, Macau residency rules tightened

  • 3

    Wild’s top line too much for Vegas Golden Knights

  • 4

    Eric Chu, 2 others to be questioned for leading illegal assembly last week

  • 5

    Taiwan draws attention at World Expo