人工智慧對法律服務及個人資料保護的衝擊

Abstract

近年來，人工智慧(Artificial Intelligence; AI）迅速發展起來。AI 各種應用及解決方案被全球的私營和公共部門組織廣泛使用，世界各國政府皆以AI做為促進經濟發展的手段， AI為個人，企業機構和社會創造廣泛而實質的利益。\n一方面人工智慧(AI）為經濟增長帶來商機，另一方面由於其本質需依賴大量的資料輸入，它對資料的使用造成潛在風險。本論文將討論AI帶來的機會及挑戰。\n技術的創新引起諸多問題的討論，包括AI是否會在法律服務行業中取代傳統的勞動力及人力律師，在構建和實施人工智慧技術時如何實際遵守個人資料保護法，以及如何解決個人資料保護原則與AI之間的衝突關係及符合現有的個人資料保護法規要求。因此，我們既有機會也有責任制定原則，最佳做法和其他問責措施，以鼓勵負責任的個人資料管理實踐，確保個人資料使用的安全，並消除不必要的障礙，以便鼓勵更多的創新技術，也重新審視現有的個人資料保護法律對AI的限制，必須確保當前的個人資料保護要求不會影響個人的隱私權或造成傷害風險。許多的報告中皆強調個人資料保護與AI應用所帶來的利益價值不是二選一而是必須找到確保兩者的方法。在第二章中，本報告介紹了人工智慧 (AI）技術，包含它是如何作用的，以及它在一些行業中的應用案例。在法律服務行業，AI 改變了服務提供的方式。人們可以通過輕鬆上網來獲得個人化合約而不是找律師擬合約。與人類律師不同，AI可以在不疲勞的情況下工作24小時不間段。它可以做重複和繁瑣的工作，如法律研究，合約審查和盡職調查，甚至可以準確有效地超越人類律師。在AI時代，人類律師必須轉變為熟悉及應用AI技術並將自己時間專注在較具創意和複雜的工作上，以提高他們的競爭力。在美國和中國，法律科技（法律AI）的應用已被廣泛使用於法律研究，合約審查和盡職調查，預測結果和機器人律師是在法律服務業中最受歡迎的領域。在台灣，受法規和市場規模的限制，法律AI並不像美國和中國那樣蓬勃發展。但是，我們在台灣看到一些新創的法律科技(法律AI) 公司。AI訓練是通過大量資料輸入來學習並隨著時間的推移擴展其知識庫。與人類律師不同，AI沒有人類特有的’心’和同理心。除了與AI技術合作外，人類律師還應培養溝通和諮詢技能等軟實力，以便與客戶建立夥伴關係。對於那些內部應用AI技術的律師事務所，在法律實踐和利用新技術的公司可以在市場上與其他公司區別開來。 AI功能高度依賴資料輸入並通過資料輸入進行學習， AI應用帶出的資料保護面臨的挑戰經常被討論。我們迫切需要詳細了解AI應用帶來的機會和問題，以及解決這些挑戰的實際方法。在第三章中，本報告討論AI應用在個人資料保護所帶來的挑戰以及可能的因應措施。其中包括：（1）資料保護原則;考慮新的做法和問責措施，列舉參考歐盟，美國和中國已經在使用的做法; （2）解釋現有資料保護法在AI應用時代不扼殺新技術的發展; （3）考慮AI應用技術及其他新技術需求的未來資料保護法。歐盟的GDPR被認為是最嚴格的個人資料保護法規。它採用了OECD(經濟合作暨發展組織) 關於保護隱私和跨境個人資料流的準則。這準則概述了資料保護，收集限制，資料質量，目的規範，使用限制，安全保障，個人參與和問責制的個人資料保護基本原則。大多數國家的個人資料保護法都遵循這些原則。在美國，AI 是一項國家議程，是提高國家競爭力的一種手段。美國政府決心成為AI的領導者。資料隱私問題在美國的AI發展戰略中被視為重要議題。 2019年3月，美國提出了一項修訂的聯邦隱私法，為消費者提供保護，以使數字經濟和創新能夠確保美國的領導作用和競爭力。在中國，AI應用是一場全國運動。中國宣布要在2020年成為AI超級大國。中國政府點名百度，騰訊，阿里巴巴，和語音識別軟件公司科大訊飛 (iFLYTEK)成為AI的全國領頭公司帶動開發AI無人駕駛汽車，疾病診斷，智能助理/機器人以及智能城市管理系統，通過各種傳感器收集資料，收集人們生活方式的資料，然後分析資料，以減少城市的環境影響，促進經濟發展和改善人的生活質量。中國的資料保護框架雖然以GDPR為藍本，但主要關注國家安全以及資料隱私與AI 和電子商務經濟增長之間的平衡。\n展望未來，本報告在最後一章節為下一步提供了建議，包括法律專業人士在AI中的新角色及與AI合作。同時促請台灣政府在AI的積極角色及重新檢視現行的個人資料保護法是否需要新的解釋，參考GDPR 在歐盟的執行情況及針對二十一世紀新建議的資料保護原則。 AI在這個階段，演算法仍然是一個“黑匣子”。雖然GDPR強制要求公司提供有關演算法背後邏輯的解釋，然而在實務中要符合這項法規將取決於技術的發展是否能做到，同時也必須考慮新設的法規是否箝制了創新的可能。在美國，2017年通過了一項新的立法來審查政府機構的”演算法偏差”。建議台灣政府可以參考GDPR 在歐盟的執行成效及美國新立法針對演算法偏差造成的歧視與資料保護相關的現行法律框架的限制，探討可能的補救措施以及確保演算法設計能符合人類的最佳利益及達到運用AI促進經濟成長的目標。

Artificial intelligence (AI) has emergently developed in recent years. Today, AI solutions are used widely by both private and public sector organizations around the globe, and governments around the world have considered AI to be a means to boost the economy. AI creates substantial benefits for individuals, institutions, and society. On one hand, AI creates business opportunities to grow the economy, and on the other hand, it causes potential risks on the data use because of its nature relying on data inputs. This paper is going o discuss the opportunities aroused by AI and potential risks exposed under it.\nThe technological innovations raise important issues including whether AI will replace the traditional workforce or the human lawyer in the legal services, how to comply with data protection laws when implementing AI technology, and how to address the challenges between AI and existing data protection requirements. Thus, we have a responsibility and an opportunity to revisit existing principles, best practices and other measures to encourage responsible data management practices and ensure the safeguard of data use and remove unnecessary roadblocks for AI’s future development. Revisiting existing data protection laws imposed on AI is essential to ensure that current data protection requirements not to harm individuals’ privacy rights. Many reports have emphasized that it cannot be one choice or another between the benefits of AI and data protection. We must find practical ways to ensure both.\nIn Chapter II, this report introduces artificial intelligence (AI) technology, how it works, and use-cases enabled by it. In the legal services, AI has changed the way of services delivery. People can access a personalized contract by going online easily rather than finding a lawyer. Unlike a human lawyer, AI can work around the o’clock without being fatigued. AI can do repetitive and tedious works such as legal data research, contract review and due diligence, and even outperform the human lawyers efficiently and cost-effectively. In the era of AI, human lawyers must transform themselves to be familiar with AI and focus on creative and complex works to enhance their competitiveness. In the US and China, legal technology (legal AI) applications have been used widely; legal data research, contract due diligence, predictions and robot lawyers are the most popular areas. In Taiwan, constrained by the regulation and the market size, legal AI is not as vibrant as it is in the US and China. However, we do see some new start-up companies in Taiwan.\nAI is trained with a high volume of data inputs. It learns through the data and expands its knowledge base over time. Unlike the human lawyer, AI does not have heart and compassion. Besides collaborating with AI, human lawyers should develop soft skills such as communication and consultative skills to develop partnerships with their clients. For those law firms developing the internal AI practice and leveraging new technologies in the legal practice are the ones who can differentiate from others on the market. AI capabilities highly rely on data inputs and learn through the data. The challenges to data protection presented by AI are frequently discussed lately. An urgent need is required for a detailed understanding of the opportunities and the issues presented by AI, and of practical ways of addressing these challenges.\nIn Chapter III, this report addresses the tensions between AI and data protection and specific responses as well as possible remedies. These include: (i) data protection principles; practices that the EU, the US, and China currently are implementing, (ii) opportunities for applying existing data protection laws to AI but not to stifle AI’s development, and (iii) considerations for future data protection laws that accommodate the demands of AI.\nThe EU’s GDPR is the strictest data protection regulations. It has adopted the OECD Guidelines on the Protection of Privacy and Transborder Flow of Personal Data. The Guidelines outline basic principles of data protection, collection limitation, data quality, purpose specification, use limitation, security safeguards, individual participation, and accountability. Most countries’ data protection laws follow these principles. In the US, AI is a top national agenda and considered to be a means to increase national competitiveness. The US government is determined to become an AI leader. The data privacy concern has been taken into consideration in its National AI Research and Development Strategic Plan. In March 2019, the US has proposed a potential revised federal privacy law providing protections for consumers and enabling the digital economy and innovation to ensure the US’ leadership role and competitiveness. In China, AI is a national movement. China has declared to be an AI superpower nation by 2020. China government has identified Baidu, Tencent, Alibaba and iFLYTEK(科大訊飛) to be national leading companies in the AI area-developing AI systems on driverless cars, diseases diagnosis, intelligent assistants/robots, and managing smart cities by collecting data through a wide variety of sensors collecting data on how people living and then analyzing that data to reduce cities` environmental impact, boost economic growth, and improve people`s lives. China’s data protection framework though has been modeled on GDPR, mainly focus on national security and a balancing act between data privacy and economic growth in AI and eCommerce.\nLooking forward, this paper provides recommendations for the next including legal professionals’ new roles in the AI era by collaborating with AI, Taiwan government’s actions on revisiting the current data protection regulation, Taiwan Personal Information Protection Act(PIPA, 台灣個人資料保護法) and building a trusted data sharing and use practice for innovation. AI algorithm remains as a ‘black box’ at this phase of AI development. Although the GDPR mandates companies to provide explanations on the logic behind the algorithms decided about individuals, the compliance will depend on the development of technologies in practice. In the US, a piece of legislation to examine ‘algorithmic bias’ in government agencies was passed in 2017. To ensure AI systems are designed for the human’s best interest and help achieve the objective of boosting the economy, Taiwan government is suggested to watch the GDPR implementation in the EU, the newly proposed data protection principles for the twenty-first century in revising the OECD’s one, and the new legislation in the US.