人工智能與不正當競爭
——揭開人工智能監管領域一個被低估的基石

斯蒂芬·舒勒（Stefan Scheuerer） 著 / 完全法律人碩士，慕尼黑馬克斯·普朗克創新與競爭研究所初級研究員

黃軍 鞠金琪 譯 / 青島大學法學院

一、 引言

長期以來，“人工智能”（簡稱“AI”）一直是知識產權和競爭法學者的關注焦點。然而，與知識產權和反壟斷不同，1. Although hinting at the Anglo-American legal sphere, the term ‘antitrust’ is preferred in this analysis over ‘competition law’ in order to avoid terminological confusion vis-a` -vis ‘unfair competition law’, since (from a European perspective) both regimes can be considered subsets of ‘competition law’, understood as an umbrella term.反不正當競爭法（簡稱“UCL”）在人工智能監管領域能夠并且應當發揮的作用迄今為止在很大程度上被忽視了。2. But see for example WIPO Conversation on Intellectual Property (IP) and Artificial Intelligence (AI), Second Session,‘Revised Issues Paper on Intellectual Property Policy and Artificial Intelligence’ (21 May 2020) para 8: ‘No separate section concerning AI and unfair competition has been added. However, recognizing that IP law and competition law clearly relate,questions have been added in the various sections (...)’.當然，反不正當競爭法是一個復雜的問題——對其本身作為一個法律領域的理解存有爭議，而在歐盟成員國之間，更不用說在世界范圍內，其在法律秩序中的體系定位和設計存在較大差異——這一事實是解釋這一缺陷的一個因素。更重要的是，這部法律體系的潛力似乎值得引起法律界人士的注意，他們對這部法律的關注還不夠深入。為了填補前述分析空白，3. Implications of AI for the legal order can be approached either from a ‘legalistic’ viewpoint, ie starting from the doctrinal framework of a specific legal regime, or from a ‘technological’/‘phenomenological’ viewpoint, ie starting from factual problems that arise in an economic, technological or societal context, cf Nicolas Petit, ‘Law and Regulation of Artificial Intelligence and robots: Conceptual Framework and Normative Implications’ (2017) 2 ＜https:/ papers.ssrn.com/sol3/papers.cfm?abstract_id?2931339＞ accessed before 27 November 2020; both approaches are important and complement each other. This articles contributes to the ‘legalistic’ dimension; for a ‘technological’ perspective, see (from an IP angle) Josef Drexl and others,‘Technical Aspects of Artificial Intelligence: An Understanding from an Intellectual Property Law Perspective’ (2019) Max Planck Institute for Innovation & Competition Research Paper No 19-13 ＜https://papers.ssrn.com/sol3/papers.cfm?ab stract_id?3465577＞ accessed before 27 November 2020.本文研究了被廣泛稱為人工智能監管的關鍵支柱和指導范式的一般原則在多大程度上反映在反不正當競爭法的特定子對應物中，從而闡明了反不正當競爭法為其成就作出貢獻的潛力。從分析角度來看，本評估過程中的一個特別重點在于從反不正當競爭法角度考慮人工智能提出的突出問題，這些問題通常在不同法律制度下被討論——以表明這種觀點可能會補充甚至取代傳統方法。在實質內容方面，將著重關注對反不正當競爭法之于人工智能創新生態系統的作用。最后，從相反的角度，本文將考慮人工智能有可能進一步助推反不正當競爭法理論體系發展的潛力，以及它對全球競爭秩序的意義。

二、場景設置：什么是人工智能，什么是反不正當競爭法，后者能對前者監管起到什么作用

人工智能與反不正當競爭法的共同之處在于，很難說它們到底是什么。人工智能是一個“包羅萬象”的術語，指涉的是圍繞大數據分析和先進算法的某些新技術，包括“自主”和“自我學習”的情形。為了在為本分析的目的揭開技術術語的神秘面紗時，機器學習（作為最重要和最突出的人工智能技術）將被視為主要參考點。4. For an overview on the technical functioning of ML and its relationship to adjacent AI technologies, see Drexl and others (n 3).反不正當競爭法是一個不那么時髦但同樣有歧義的現象：它在國際層面首先體現在1803年 《保護工業產權巴黎公約》第十條之二，其歷來被視為在競爭中保護“倫理”或者“商業倫理”，依靠“尊貴商人”的理想模式。現代學界通過運用功能經濟學的維度來構建反不正當競爭法，其假定與反壟斷法的最終互補性，并將保護競爭作為一項制度的中心目標。5. cf Reto M Hilty, ‘The Law Against Unfair Competition and its Interfaces’ in Reto M Hilty and Frauke Henning-Bodewig(eds), Law Against Unfair Competition - Towards a New Paradigm in Europe? (Springer 2007) 1; Rupprecht Podszun, ‘Der‘more economic approach’ im Lauterkeitsrecht’ [2009] WRP 509.盡管如此，反不正當競爭法規則的準確設計和理解在歐盟成員國和全世界范圍內都有相當大的差異：從競爭法的編纂到消費者法、公法；從實質上講，在保護競爭、消費者和作為一個競爭制度之間搖擺不定。6. For an overview, see Frauke Henning-Bodewig, International Handbook Of Unfair Competition (CH Beck/Hart/Nomos 2013);illustrative of the scattered nature, Richard Arnold, ‘English Unfair Competition Law’ (2013) 44 IIC 63, 77: ‘It is still the case that English law does not recognise any general tort of unfair competition. It does not follow, however, that there is no English law of unfair competition’; on the difficulties of determining UCL, see also Frauke Henning-Bodewig and Achim Spengler,‘Conference Report: “Framing - The ‘Hard Core’ of Unfair Competition Law”’ [2016] GRUR Int 911.盡管歐盟通過《不正當商業行為（UCP）指令》對反不正當競爭法中的企業對消費者（B2C）層面進行了協調，7. Directive 2005/29/EC of the European Parliament and of the Council concerning unfair business-to-consumer commercial practices in the internal market.但企業對企業（B2B）層面迄今為止尚未被統一。8. As far as the B2C dimension is concerned, this article will focus on European law; as far as the B2B dimension is concerned,on German law as an illustrative and doctrinally advanced example or blueprint.

然而，這種模糊性并不一定是反不正當競爭法潛在運用于人工智能監管領域的不利條件。誠然，鑒于上述分歧，它在協調監管方面幾乎不會立即產生效果。然而，首先“監管競爭”的理念可能會帶來收益。尤其是歐盟層面的B2B領域反不正當競爭法不協調的事實，從這個角度來看，其應被視為機遇。人工監管領域與其尋求監管的技術同樣是動態的。至于如何對待反不正當競爭法視域下的人工智能，各國相互競爭的措施本身可能被視為“監管沙箱”：9. On regulatory sandboxes for data sharing, cf Rupprecht Podszun, ‘Datenpools: Ausprobieren statt differenzieren‘[2019] WUW 289.找到的最佳解決方案可以出口到其他司法管轄區——無論是在立法層面，還是在通過比較法律方法對一般條款進行司法解釋的層面。第二，在相關方面，反不正當競爭法固有的特殊靈活性十分契合人工智能領域的動態屬性，它將對全部法律秩序的理解最終進行統一。反不正當競爭法可以作為一種“后備”機制發揮可行的作用，在缺乏具體立法情況下以應對新的和不可預見的競爭風險。這種后備屬性屬于反不正當競爭法的傳統特征，它為從理論發展到后來明確的法典化創造了肥沃的土壤。10. cf Herbert Zech, Information als Schutzgegenstand (Mohr Siebeck 2012) 161 f.它在數字經濟中獲得了更大的意義。

當前人工智能監管原則在反不正當競爭法范式中的體現程度如何？關于人工智能監管框架的爭論是動態的、持續的，現在談論是一個成熟的知識顯然還為時過早。盡管如此，在學術探討和公共及私人機構的眾多政策指南中，可以找到總體上且反復出現的相關范式的某種共識。在反復被援引的原則中，包括全面實現“道德”、公平、透明、問責、自主和促進創新。11. cf only High-Level Expert Group on Artificial Intelligence, ‘Ethics guidelines for trustworthy AI’ (2019) ＜https://ec.europa.eu/digital-singlemarket/en/news/ethics-guidelines-trustworthy-ai＞ ccessed before 27 November 2020; OECD, ‘Council Recommendation on Artificial Intelligence’ (2019) ＜https://legalinstruments.oecd.org/en/instruments/ OECD-LEGAL-0449＞accessed before 27 November 2020; this list of values is by no means exhaustive, yet these appear to be the most prominent ones.以下考慮事項將闡明，反不正當競爭法如何具體有助于實現這些目標。

三、“人工智能倫理”和“商業倫理”：監管嵌合體的(非)融合

首先，人們通常可以思考，人們所廣泛宣揚的“人工智能倫理”12. cf High-Level Expert Group (n 11); IEEE, ‘Ethically Aligned Design - A Vision for Prioritzing Human Well-being with Autonomous and Intelligent Systems’ (2019) ＜https://standards.ieee.org/content/dam/ieeestandards/standards/web/documents/other/ead1e.pdf?utm_medium? undefined&utm_source?undefined&utm_campaign?undefined&utm_content?undefined&utm_term?undefined＞ accessed before 27 November 2020.的愿景與“商業倫理”的概念之間是否存在聯系，這種聯系通常或至少在歷史上與反不正當競爭法有關。這顯然觸及了關于反不正當競爭法到底是什么的爭論內核。正如前文所述，在其歷史根源上曾經是一個涉及競爭“倫理”的法律領域。13. It is worth noting, however, that now as before, irrespective of the ‘moral’ rhetoric and underpinnings, the practical application of the law has often followed a functional balancing of interests.盡管這種理解在很大程度上被現代經濟功能方法所取代，但舊的理解碎片仍然滲透在法律、判決和學術探討之中，各成員國的側重點也各不相同。如果有人認為“商業倫理”在法律秩序中仍有一席之地，而該領域就是反不正當競爭法，那么將相關原則與“人工智能倫理”的要求結合起來似乎并不牽強。然而，本文的立場并非宣揚這一主張，而是要指出迫切需要對“倫理”敘事進行去神秘化。首先也是最重要的是，如果缺少“法律”的鏡像，就很難有“倫理”價值觀，尤其是與各自價值觀有關的基本權利或人權，14. cf High-Level Expert Group (n 11) 37, however, considering fundamental rights a mere sub-realisation of ethics.這使得“倫理”的整個概念更令人困惑，而不是有助于實現法學研究目的。其次，通常被視為“不道德”的行為往往與反競爭行為具有一致性。在任何情況下，顯然只有“人工智能倫理”中與市場和競爭相關或影響市場和競爭的部分才與反不正當競爭法相關。最后，當涉及具體法律運作時，所有這些問題，無論其形而上學的起源如何，均可歸結為所有市場參與者合法利益的平衡。這種平衡是反不正當競爭法理論的核心。因此，以下考慮將包含法律而非“倫理”反思。

四、公平

人工智能和反不正當競爭法最明顯、同時也是最復雜的潛在“共同點”是“公平”原則本身。從表面上看，反不正當競爭法中的“公平”和人工智能語境中的“公平”可能被認為除了術語之外并無共同之處：人工智能爭論中的“公平”大多被理解為平等原則和禁止“有偏見”的歧視，反不正當競爭法中“公平”旨在保護競爭或至少與競爭相關的利益。15. Of course, some phenomena of ‘discrimination’ have immediate competitive relevance, for example the prohibition imposed on dominant companies not to apply dissimilar conditions under art 102(c) TFEU; on the connection between anti-discrimination legislation and UCL, see also section VIII.2. below.然而，這兩個概念不僅具有內在的開放性和模糊性。16. cf High-Level Expert Group (n 11) 12: ‘(...) we acknowledge that there are many different interpretations of fairness (...)’.人們也不應忽視人工智能的（錯誤）使用可能帶來的諸多負面影響，尤其是對競爭的負面影響。雖然這主要體現在反壟斷場景中，如“算法共謀”，17. cf only Mark-Oliver Mackenrodt and Francisco Beneke, ‘Artificial Intelligence and Collusion’ (2019) 50 IIC 109.但人工智能影響的領域往往與傳統上的反不正當競爭法相關，尤其是與“消費者保護”的相關領域。下面將提供示例。盡管禁止“不正當”商業行為的反不正當競爭法一般條款的顯著特性固然可以解決新的和不可預見的競爭風險，但基于邏輯原因，本文沒有對此進行進一步闡述。

雖然這不是深入探討“公平”的實質意義（或者更確切地說：它所包含的多重維度）的持續和長期爭論的地方，但反不正當競爭法對“公平”市場秩序的一個非常具體的貢獻值得強調：它與反壟斷法的監管互補性。從實質上講，反不正當競爭法可以解決未能達到市場支配地位反壟斷要求的競爭問題。18. cf Heike Schweitzer and others, ‘Modernisierung der Missbrauchsaufsicht fu¨ r marktma¨ chtige Unternehmen’(2018)107,110＜https://www.bmwi.de/Redaktion/DE/Publikationen/Wirtschaft/modern isierung-der-missbrauchsaufsichtfuer-marktmaechtige-unternehmen. html＞ accessed before 27 November 2020; Peter Picht and Gaspare Loderer, ‘Framing Algorithms - Competition Law and (Other) Regulatory Tools’ (2018) Max Planck Institute for Innovation & Competition Research Paper No 18-24, 33 ＜https://papers.ssrn.com/ sol3/papers.cfm?abstract_id?3275198＞ accessed before 27 November 2020: ‘Although not addressed in detail here, rules against unfair competition are another major element, both as a template for and a tool complementary to the provisions against cartels and abuse of dominance’; on data access as a concrete example where this complementarity becomes relevant, see section IX.1. below.鑒于在數據驅動型市場中確定市場力量的難度，其重要性愈加凸顯。19. cf Boris Paal and Moritz Hennemann, ‘Big Data im Recht’ [2017] NJW 1697, 1699.當然，考慮到理論的體系性，必須謹慎，不要繞過或破壞反壟斷法的結論性決定，即非支配主體采取的某些行為在反不正當競爭法中并不具有違法性。然而，如果人們依循對反不正當競爭法的“現代”理解，將保護競爭作為一項機制置于其目的的關注中心，那么其一般條款可以作為解決反壟斷領域之外人工智能所引致的市場失靈的基石。

五、透明

透明是人工智能監管的核心準則。人們普遍希望人工智能本身的解決（與純人類決策相反）和人工智能實現決策的具體方式（通常被稱為“黑箱”問題，通過努力實現“可解釋的人工智能”來反映）均是透明的。20. Of course, the feasibility of transparency in the latter regard ultimately depends on the technological state of the art, cf Deven Desai and Joshua Kroll, ‘Trust But Verify - A Guide to Algorithms and the Law’ (2017) 31 Harvard Journal of Law &Technology 1.當前透明有多種表現形式，但一個重要的形式無疑是市場透明度。保護市場透明度的傳統體系領域是反不正當競爭法，其禁止誤導性商業行為。21. cf Section 1, arts 6 and 7 UCP Directive.在各自理論測試下，最終具有決定性是一項決策的起源是基于算法還是人為的，以及這是否會影響消費者的商業決策。22. cf Benjamin Raue and Antje von Ungern-Sternberg, ‘Ethische und rechtliche Grundsa¨ tze der Datenverwendung‘[2020]ZRP 49, 52.

1.基于人工智能的個性化

人工智能應用的主要和最具經濟價值的領域之一是用于個性化策略，尤其是個性化定價和個性化廣告。23. On personalised advertising see Guido Noto La Diega, ‘Data as Digital Assets. The Case of Targeted Advertising’ in Mor Bakhoum and others (eds), Personal Data in Competition, Consumer Protection and Intellectual Property Law - Towards a Holistic Approach? (Springer 2018) 447. The extent to which such personalisation actually happens in practice remains dubious, cf OECD Secretariat, ‘Personalised Pricing in the Digital Era’ (2018) ＜https://one.oecd.org/document/DAF/COMP/WD(2018)146/en/pdf＞ accessed before 27 November 2020; further empirical research is needed in this area.圍繞是否應當禁止或限制這種個性化策略展開了激烈辯論，即使這些策略是提高整體福利的，理據在于消費者普遍認為它們是“不正當的”或者“不公正的”。24. See on this debate Christopher Townley, Eric Morrison and Karen Yeung, ‘Big Data and Personalised Price Discrimination in EU Competition law’ (2017) King’s College London Dickson Poon School of Law Research Paper No 2017-38 ＜https://papers.ssrn.com/sol3/papers. cfm?abstract_id?3048688＞ accessed before 27 November 2020; Gerhard Wagner and Horst Eidenmu¨ ller, ‘Down by Algorithms? Siphoning Rents, Exploiting Biases, and Shaping Preferences: Regulating the Dark Side of Personalized Transactions’ (2019) 86 University Of Chicago Law Review 581, 587.在不深入討論情形下，有一件事似乎是無可爭議的：消費者必須知道他或她受制于個性化策略，而未得到同等對待。25. On the respective regulatory potential of UCL, see Picht and Loderer (n 18) 33; Wagner and Eidenmu¨ ller (n 24) 590; cf also the precontractual information duty on personalised pricing on the basis of automated decision-making according to art 4(4)(a)(ii) Directive (EU) 2019/ 2161.在某種程度上，如果消費者不是基于自主和知情的決策行事，個性化可能因此違反反不正當競爭法規定的透明度規則。26. cf Helga Zander-Hayat, Lucia Reisch and Christine Steffen, ‘Personalisierte Preise - eine verbraucherpolitische Einordnung’[2016]VuR 407; Franz Hofmann, ‘Der ma?geschneiderte Preis - Dynamische und individuelle Preise aus lauterkeitsrechtlicher Sicht’ [2016] WRP 1080; on the implications for consumer ‘autonomy’, see section VII. below.尤其是價格透明度的缺失會造成信息不對稱，從而消除了對競爭至關重要的價格比較可能性，進而損害經濟福利。27. cf Zander-Hayat, Reisch and Steffen (n 26) 407 f.當然，精確的信息要求也有爭議：為了不引起“信息過載”，必須對其加以平衡，28. In this regard, personalisation offers interesting possibilities: each consumer could get personalised information, exactly suiting his or her capabilities, situation and needs. Ultimately, this is one aspect of what is currently discussed under the vision of ‘personalised law’, cf (critically) Philip Bender, ‘Limits of Personalization of Default Rules - Towards a Normative Theory’(2020) Working Paper of the Max Planck Institute for Tax Law and Public Finance No 2020-02 ＜https://papers.ssrn.com/ sol3/papers.cfm?abstract_id?3544029＞ accessed before 27 November 2020.并且必須根據在特定社會或商業環境中合理預期或常見（包括人工智能的使用程度）來解釋，29. According to Hans-Wolfgang Micklitz and Monika Namyslowska, ‘§ 5a UWG’ in Gerhard Spindler and Fabian Schuster(eds), Recht der elektronischen Medien (4th edn, CH Beck 2019), average consumers don’t expect prices to be personalised, and this circumstance may be relevant for their commercial decision. Conversely, should personalised pricing develop to become such common practice that the public generally expects to be subjected to personalised prices as the new digital normal, the misleading character would vanish.以及對于所討論的商業傳播媒介，哪些履行信息義務的方式和方法是適當的。30. Considerations de lege ferenda include a duty to highlight the use of AI via visual symbols, cf Martin Ebers, ‘Ku¨ nstliche Intelligenz und Verbraucherschutz’ [2020] VuR 121.

2.人工智能相關營銷的進一步示例

除了個性化，反不正當競爭法還可以解決人工智能相關營銷活動的透明度問題。首先，鑒于“人工智能”術語的模糊性，人們可以考慮在“AI”一個誤導性實踐的誘人承諾下，對“正常”計算機軟件的營銷行為進行討論。其次，企業越來越多地公布與人工智能相關的行為準則，在這些準則中，其或多或少的具體說明了他們打算如何使用人工智能來造福社會，并避免歡迎的行為。31. cf Google, ‘AI at Google: our principles’ (7J une 2018) ＜https://www. blog.google/technology/ai/ai-principl es/＞ accessed before 27 November 2020.這些準則可被視為“企業數字責任”現象的一部分，即“企業社會責任”的數字化延續。32. On the latter, see comprehensively Reto M Hilty and Frauke Henning-Bodewig (eds), Corporate Social Responsibility.Verbindliche Standards des Wettbewerbsrechts? (Springer 2014).

如果一家公司違反了該準則中的聲明，反不正當競爭法將在打擊欺騙行為和恢復市場透明度方面發揮重要作用。33. cf Frauke Henning-Bodewig, ‘TRIPS and Corporate Social Responsibility: Unethical Equals Unfair Business Practices?’ in Hanns Ullrich and others (eds), TRIPS plus 20 (Springer 2016) 701, 714.因為，如果公司想利用他們的“良好表現”作為針對重視此種表現的消費者的競爭優勢，那么唯有在所做的宣傳得到切實履行的情況下，基于這些理由的競爭才能發揮作用。在這方面，法律適用的主要問題是許多聲明的模糊性。34. cf ibid.例如，人們很難從諸如以“有益社會”35. Google (n 31) principle No 1.方式使用人工智能的承諾中得出結論。

再次，另一組可能越來越具有相關性的案例來自知識產權法領域，并且涉及區分無形商品尤其是在版權意義上看起來像“作品”的無形商品是由人類創造的，還是在人工智能的大力幫助下創造的必要性。對于“人工智能生成”作品的知識產權保護的正當性而言，大量的人工指導是否是必要的，這一問題一直且將進一步讓知識產權學者忙碌。36. cf in more detail section IX.2. below.然而，可以肯定的是，在法律上討論“人類制造”和“人工智能生成”的區別面臨著一個現實挑戰，即必須辨別各自的起源。其中的一個市場解決方案有賴于消費者對于人工生成作品的評價，而不是人工智能生成的作品，如果實際上無法區分彼此，則不起作用。37. On this ‘market solution’, see in more detail Reto M Hilty, Jo¨ rg Hoffmann and Stefan Scheuerer, ‘Intellectual Property Justification for Artificial Intelligence’ (2020) Max Planck Institute for Innovation & Competition Research Paper No 20-02, 11＜https://papers.ssrn.com/ sol3/papers.cfm?abstract_id?3539406＞ accessed before 27 November 2020.如果AI生成的“作品”被當作人造產品進行營銷，這種營銷無論是主動還是被動地隱瞞人工智能的來源，都可能構成不正當競爭行為：一種誤導行為。38. In this regard, it should be noted that the potential non-registrability of subject matter generated ‘autonomously’ by AI may also lead companies to conceal the use of AI when registering their inventions or designs, cf WIPO (n 2) para (vii); Sven Hetmank and Anne Lauber-Ro¨ nsberg, ‘Ku¨ nstliche Intelligenz - Herausforderungen fu¨ r das Immaterialgu¨ terrecht’ [2018]GRUR 574, 581 therefore suggest that a labelling requirement as to AI involvement could be introduced as a protection criterion for AI-generated products to establish transparency; on deceptive conduct before patent offices and potential remedies, see generally Eugenio Hoss, Deceptive Conducts before the Patent Office (Nomos 2019).

六、責任

確保企業對其人工智能造成的“自動”損害負責是與人工智能相關的最“經典”的法律問題。39. An aspect worth highlighting in this context is the overestimation of the relevance of ‘autonomy’ notions: in many cases, it is simply decisive whether there has been (in)sufficient guidance of foreseeability of certain AI-induced results ‘on the human side’, irrespective of the ‘autonomy’ degree ‘on the AI side’.最典型的例子是自動駕駛汽車碾壓行人。然而，AI也可能“自主地”損害知識產權或總體競爭。在反不正當競爭法的案例中，關鍵問題在于確定是由或者在企業人工智能幫助下實施的不公平商業行為的責任。在近期的學術討論之中，人們便強調了對此類“歸因問題”的整體概念的需要，即將圍繞“次要責任”等概念的有點支離破碎的理論框架整合到一個連貫的框架中。40. cf Franz Hofmann, ‘Disziplinarita¨ t, Intradisziplinarita¨ t und Interdisziplinarita¨ t am Beispiel der Grundsa¨ tze “mittelbarer Verantwortlichkeit”’ [2018] JZ 746; on harmonisation possibilities of intermediary liability cf Matthias Leistner, ‘Intermediary Liability in a Global World’ in Tatiana Eleni Synodinou (ed), Pluralism or Universalism in International Copyright Law (Kluwer Law 2019)當具體從人工智能角度來構建這樣的框架時，在已有經驗基礎上構建框架，即在不同法律制度的責任歸因領域已經形塑的范例，而不是從零開始創造全新的概念，這種做法似乎是明智的。反不正當競爭法可以是這些理論上鼓舞人心的制度之一。

在德國，“競爭中違反注意義務的責任”的概念是根據反不正當競爭法一般條款發展而來的，從而作為知識產權法“侵權責任”的一種替代方案。41. cf German Federal Supreme Court, 12 July 2007, I ZR 18/04 - Jugendgefa¨ hrdende Medien bei ebay.它提供了將反競爭行為歸責于公司的理論指導，使得公司對未履行其阻止相關行為的職責負責。該范式特別是在或者為了互聯網群加以發展和改進，特別需要對相應措施的范圍和“合理性”制定標準，包括它們在多大程度上包含防止未來相同或類似侵權行為的義務。42. cf Ansgar Ohly, ‘§ 8 UWG’ in Ansgar Ohly and Olaf Sosnitza (eds), Gesetz gegen den unlauteren Wettbewerb (7th edn, CH Beck 2016) para 127.已經有人提出將這一概念作為一種潛在的合理模型移轉于人工智能情形引發的違反反壟斷法行為的情形。43. cf Moritz Hennemann, ‘Ku¨ nstliche Intelligenz und Wettbewerbsrecht’ [2018] ZWeR 161, 180 f.它可能為歸因分歧提供有價值的架構、先例和參考因素，提升法律確定性，并在商業自由和防止損害之間實現經濟上的合理權衡。最后，鑒于充分責任和創新之間的聯系，這一問題也可以被視為反不正當競爭法對促進人工智能創新的貢獻，這一目標將在下文第九部分進一步闡述。

七、自主性

如果我們遵循科幻小說啟發的概念，人工智能的最終威脅是其取代人類的潛力。然而，保護人類自主性是人工智能監管原則的核心。44. cf only High-Level Expert Group (n 11) 12; OECD (n 11) IV.1.2.a).反不正當競爭法建立在并致力于維護人類經濟的一個非常重要的子方面：消費者作為市場參與者的自主性，他們讓競爭的概念在履行其“仲裁員角色”時發揮作用。人工智能在此方面提出了兩方面的問題。

1.供應商使用人工智能的自主性威脅

人工智能在供應端的使用，尤其是在個性化策略中的使用，將新產品和廣告完全基于既有偏好，可能會在“過濾泡沫”中捕獲消費者。在這種偏好定制系統的擴散過程中，來自各種市場選項的自主性選擇可能會消減。然而，好消息是，反不正當競爭法大體上提供了解決這些威脅的方法：如前所述，透明度要求至少緩解了緊張情形。45. cf section V.1. above; Wagner and Eidenmu¨ ller (n 24) 590 on personalised pricing: ‘An obligation to disclose the application of first-degree price discrimination appears innocuous and potentially effective to leverage consumer autonomy.’消費者自愿或非自愿地進入或停留在過濾泡沫之中，這是一種自主選擇，盡管自愿喪失自我能力的悖論和危險人所周知。在這種情形下，反不正當競爭法似乎是對抗過度“過濾泡沫”問題的以競爭為導向的子支柱。46. A parallel problem regarding ‘filter bubbles of opinion’ threatening democracy is debated in the media law realm, cf Josef Drexl, ‘Bedrohung der Meinungsvielfalt durch Algorithmen’ [2017] ZUM 529.

2.消費者使用人工智能的自主性威脅

（甚至）更成問題的是，消費者使用人工智能的鏡像維度，特別是在依賴物聯網（以下簡稱“IoT”）應用時，“算法消費者”一詞已經被創造出來。47. Niva Elkin-Koren and Michal Gal, ‘Algorithmic Consumers’ (2017) 30 Harvard Journal of Law & Technology 309.一個例子是“智能家居”中的“自動冰箱”，它可以在沒有人類消費者（主動）參與的情況下（根據先前的偏好）訂購新的食物。一方面，這種使用可能構成一種受歡迎的“以牙還牙”對抗策略，以對抗企業對人工智能的損害，恢復技術和信息的平衡，同時，從人類學的角度來看，它可能會剝奪消費者作為理性市場代理人的能力，因為他們的所有決策均由人工智能工具來完成。48. cf the concerns articulated by Josef Drexl at the ‘Consumer Law Days 2019’ conference, reported by Jure Globocnik and Stefan Scheuerer, ‘Datenzugang, Verbraucherinteressen und Gemeinwohl - Bericht u¨ ber die Verbraucherrechtstage 2019 des Bundesministeriums der Justiz und fu¨ r Verbraucherschutz in Berlin’ (2020) 11 JIPITEC 228, 229.

關于反不正當競爭法應對自身基礎威脅的潛在解決方案，在理論上接受并調整“算法消費者”，尤其是“一般消費者”標準的構建似乎是必要的，49. The ‘average consumer’ standard, against which misleading practices are judged, is not only challenged by personalisation phenomena that question the very concept of ‘average’ (cf Peter Rott, ‘Der “Durchschnittsverbraucher” - ein Auslaufmodell angesichts personalisierten Marketings?’ [2015] VuR 163). Also, with a view to ‘algorithmic consumers’, a ‘technicised’reconstruction of this hypothetical figure as ‘average algorithmic consumer’ may become necessary.但不足以解決自主性問題。相反，如同其他法律領域一樣，很可能有必要“讓人參與其中”。例如，冰箱可能會被迫不時查看消費者，詢問他們的偏好是否發生了變化，或者是否對新的報價感興趣。這樣的義務通常必須在反不正當競爭法之外實現。

盡管如此，反不正當競爭法借助其在消費者選擇模式方面的豐富經驗，可以為決策者提供理論指導，以評估有多少決策權可以委托給“算法消費者”，以及有多少決策權不能委托給“算法消費者”，而不破壞市場秩序本身的功能。尤其是，反不正當競爭法理論可以在此方面影響關于通過設計實現各自規范的爭論。50. cf IEEE (n 12).

八、反不正當競爭法作為與人工智能相關的特別反不正當競爭法市場行為規則的實施工具

（一）在執法環境中定位反不正當競爭法

反不正當競爭法可以作為反不正當競爭法以外各類市場行為規則的（額外）執行支柱，違反這些規則會通過“違反法定義務”這一原則對競爭產生負面影響。在程序方面，這一選擇通過競爭對手和消費者協會釋放了執法的可能性，許多法律秩序有賴于反不正當競爭法，從而在與反壟斷法相關的國家當局之外提供了一種制度性補充。冗長的行政程序相比，這種執法方式速度更快、更靈活，因此顯示出特別適合人工智能和數字經濟的特征。從實質上講，“違反法定義務”似乎是一個恰當的理論工具，可以將正在進行的有關數字經濟中保護消費者利益的法律領域日益趨同的討論付諸實施。在這些機制下可能受到制裁的眾多違法行為中，有三種似乎與人工智能背景尤為相關：歧視、個人數據保護和網絡安全。51. The phenomena of discrimination and personal data protection can be seen in conjunction with the personalisation problem outlined above, as personalisation can be based on data gathering in violation of data protection rules, and if the personalisation relies on traits protected by anti-discrimination laws, it may also violate the latter.

（二）市場相關環境下基于人工智能的歧視

反歧視立法是從法律上判斷“人工智能偏見”問題的標準。盡管反歧視規則不是與市場行為相關的規則，但在反不正當競爭法范域內，它們可以在某些情況下適用。一個明顯的例子是上述商業環境中的個性化策略，即如果個性化是基于反歧視法禁止提及的特征，如種族或性別。盡管這些方面在一開始就以非經濟價值為基礎，如人的尊嚴和個性，但它們仍然影響和限制著企業在市場上的行為表現。

（三）競爭與隱私：是敵是友?

AI對社會造成的最基本與最具體52. As opposed to the more far-reaching sci-fi dystopias circling around the discourse.的威脅在于它有能力建立全方位的監控，包括國家和私營企業的監控。53. cf Marc Amstutz, ‘Dateneigentum’ (2018) 218 AcP 438, 520, diagnosing the threat of ‘algorithmic governmentality’ based on big data gathering (although not considering data protection laws the correct or sufficient remedy).因此，將強有力的數據保護規則與競爭規則所追求的以市場和福利為導向的經濟目標保持一致至關重要。54. In this regard it is worth highlighting that both welfare and data protection are collective societal interests, cf Indra Spiecker genannt Do¨ hmann at the Consumer Law Days 2019 (n 48) 233.德國競爭主管機構聯邦反壟斷局對Facebook的調查引發了關于競爭法和數據保護法之間關系的激烈辯論，該機構將濫用主要基于支配地位的行為視為違反數據保護規則。55. Bundeskartellamt, 6 February 2019, B6-22/16; German Federal Supreme Court, 23 June 2020, KVR 69/19 - Facebook II; see on the respective discussion Marco Botta and Klaus Wiedemann, ‘The Interaction of EU Competition, Consumer, and Data Protection Law in the Digital Economy: The Regulatory Dilemma in the Facebook Odyssey’ (2019) 64 The Antitrust Bulletin 428; Klaus Wiedemann, ‘A Matter of Choice: The German Federal Supreme Court’s Interim Decision in the AbuseofDominance Proceedings Bundeskartellamt v. Facebook (Case KVR 69/ 19)’ (2020) 51 IIC 1168.與此同時，還有一種討論是關于違反數據保護是否可以作為違反反不正當競爭法的法定義務予以制裁。56. cf Ansgar Ohly, ‘UWG-Rechtsschutz bei Versto¨ ?en gegen die Datenschutz-Grundverordnung?’ [2019] GRUR 686.如果人們遵循上述反壟斷法和反不正當競爭法之間目的互補的觀點，認可兩部法律本質上均是為了保護有效競爭（或最大化福利）的同一目標，然后將這兩條討論進路聯系并結合起來似乎至關重要。57. cf Torsten Ko¨ rber, ‘Die Facebook-Entscheidung des Bundeskartellamtes’ [2019] NZKart 187, considering the Facebook proceedings an antitrust equivalent to UCL’s breach of statutory duty doctrine.從反壟斷的角度來看，檢驗的是違反數據保護規則的行為是否可以被視為屬于市場主導行為人“利用”客戶或“阻礙”競爭對手的既定類別；在反不正當競爭法語境下，需要違反市場行為規則，并對市場參與者的（與競爭相關的）利益產生相當大的影響。然而，這兩個方面的共同問題似乎是，數據保護規則在多大程度上與競爭有內在聯系，或者因違反數據保護法而對競爭造成損害需要哪些特定于競爭的“額外條件”。58. As regards breach of statutory duty in the EU, the discussion is overlapped by the systematic issue of whether the GDPR sanction regime is conclusive and thus prevents relying on additional enforcement mechanisms. This question is out of the scope of this paper, as it gives no guidance on the substantive relationship between data protection law and competition law.

這個問題的答案很復雜，而且反思仍在繼續。然而，本文希望強調一些理論指引：首先，將“隱私”理解為一種經濟商品并將其納入經濟福利理論的努力需要進一步追求和推進。59. Welfare theory is ultimately about the (pareto-)optimal allocation of goods: if privacy can be understood as a good that has to be optimally allocated, it may well be included in an overall welfare doctrine spanning both competition and data protection law; on the economics of privacy, see Alessandro Acquisti, Curtis Taylor and Liad Wagman, ‘The Economics of Privacy’ (2016)54 Journal of Economic Literature 442; pessimistic, Bertin Martens at the Consumer Law Days 2019 (n 48) 231, considering the economic value of privacy still being insufficiently understood and economics thus being of little help for balancing welfare with data protection interests; optimistic, Ryan Calo, ‘Privacy and Markets: A Love Story’ (2016) 91 Notre Dame Law Review 649.這樣，隱私作為數字經濟的核心消費者利益，最終可能會被視為“消費者福利”的組成部分。遵循一種普遍思路，這是競爭法應當追求的規范性標準，同時也被認為是需要重構和適應數字時代的標準。60. cf European Data Protection Supervisor, ‘Preliminary Opinion: Privacy and competitiveness in the age of big data: The interplay between data protection, competition law and consumer protection in the Digital Economy’ (March 2014) para 71＜https://edps.europa.eu/sites/edp/files/ publication/14-03-26_competitition_law_big_data_en.pdf＞ accessed before 27 November 2020: ‘Given the reach and dynamic growth in online services, it may therefore be necessary to develop a concept of consumer harm, particularly through violation of rights to data protection, for competition enforcement in digital sectors of the economy’.其次，關于隱私/人格61. There are complex differentiations regarding the concepts of ‘privacy’ and ‘personality’ and their interrelation. Elaborating on these lies beyond the scope of this article.和知識產權概念上重疊的理論知識應納入討論：盡管這兩種制度的側重點不同，但都作為無形標的物權利的基礎被納入了基于經濟和人格本位的理由中，同時針對知識產權與競爭法之間關系的理解似乎遠比隱私與競爭法之間更為進步。62. On the relationship between privacy and intellectual property, cf Diana Liebenau, ‘What Intellectual Property Can Learn from Informational Privacy, and Vice Versa’ (2016) 30 Harvard Journal of Law and Technology 285; on a historical side note,it seems illustrative to recall that the influential German scholar Josef Kohler once considered the whole body of (B2B) UCL as protecting the ‘personality interests’ of companies, cf Josef Kohler, Der unlautere Wettbewerb (Rothschild 1914) 17 ff; one can still reflect on whether to locate in particular trade secrecy interests purely in the realm of economics, to view them from an IP angle, or to theorise them in conjunction with privacy and ‘corporate personality’ paradigms.第三，無論如何，這種反思的結果很可能是數據保護規則的混合屬性，包括一些可以適用于經濟范式的要素和其他不能適用于經濟范式的要素。63. cf the differentiation by Francisco Costa-Cabral and Orla Lynskey, ‘The Internal and External Constraints of Data Protection on Competition Law in the EU’ (2015) LSE Working Papers 25/2015, 3, assuming that ‘EU data protection norms may impose both an internal and an external constraint on the application of competition law’.第四，也是最后一點，盡管存在目的多元化和重疊性，但必須注意一條基本的系統分界線：在不損害競爭情況下，數據保護不能也不應通過競爭機制純粹以“執法協助”為由來實施。64. Against an expansionist use of breach of statutory duty in non-market related contexts, see generally Ansgar Ohly, ‘§3a UWG’ in Ansgar Ohly and Olaf Sosnitza (n 42) para 21; anyway, the rather strong enforcement regime of the GDPR has mitigated the need of externally assisting the formerly ‘toothless tiger’ data protection law.

（四）網絡安全

網絡安全對于人工智能和物聯網生態系統的運行和信譽至關重要。需要防止黑客入侵的“自動駕駛汽車”再次提供了一個例證。雖然網絡安全的法律理論仍處于起步階段，但其作為一套市場行為規則的性質似乎是無可爭議的。65. For a comparative law overview on the legal framework, cf DennisKenji Kipker and Sven Mueller, ‘International Regulation of Cybersecurity - Legal and Technical Requirements’ [2019] MMRAktuell 414291.如果違反此類規則，反不正當競爭法規定的責任可以作為（額外）誘因，促使企業充分維護各自的標準。66. cf Thomas Riehm and Stanislaus Meier, ‘Rechtliche Durchsetzung von Anforderungen an die IT-Sicherheit’ [2020] MMR 571, 574 f.

九、反不正當競爭法對促進人工智能創新監管框架的貢獻

人工智能的另一個關鍵承諾是促進創新。反不正當競爭法至少可在三個方面為促進創新法律框架作出貢獻，以下概述之。

（一）反不正當競爭法背景下的數據訪問

數據訪問是人工智能創新的關鍵之所在。特別是“機器學習”嚴重依賴數據。過去幾年里，關于獲取這些數據的爭論已經取得了很大進展。67. Numerous possible doctrinal foundations de lege lata and de lege ferenda have been invoked for granting such access; for a comprehensive overview, cf Federal Ministry of Justice and Consumer Protection and Max Planck Institute for Innovation and Competition (eds), Data Access, Consumer Interests and Public Welfare (Conference Volume on the Consumer Law Days 2019)(forthcoming).然而，在討論中很少考慮從反不正當競爭法領域推斷數據訪問機制的選項。68. See the proposal of Drexl at the Consumer Law Days 2019 (n 48) 237 and 238; in detail, Josef Drexl, ‘Connected Devices- An Unfair Competition Law Approach to Data Access Rights of Users’ (2020) Max Planck Institute for Innovation and Competition Research Paper No 20- 22 ＜https://papers.ssrn.com/sol3/papers.cfm?abstract_id?3746163＞ accessed before 31 December 2020.如果數據訪問利益可以定位于傳統上與反不正當競爭法的相關領域，那么出于體系一致性的原因，它們應該位于那里而非其他領域。69. On the value and necessity of locating claims in the fitting legal regime, see the discussion at the Consumer Law Days 2019(n 48) 238; from the perspective of applicable law, Drexl (n 68) 42.然而除此之外，還應探究反不正當競爭法作為競爭相關問題的創新“蓄水池”的潛力，對于這些問題，其他任何系統領域都不是直觀的、顯眼的或在眼前的。70. On this feature of UCL, see section II. above.反不正當競爭法進路可以解決與B2B和B2C維度相關的訪問問題，而且似乎不需要明確的法律提案，71. But see in this vein Drexl (n 48) 237.盡管這肯定利于法律的明確性。相反，就目前而言，反不正當競爭法一般條款可以勝任此項任務。

首先，橫向請求可能導致“故意阻礙競爭對手”的后果，德國反不正當競爭法針對B2B行為設置了“小一般條款”，其基于整體利益平衡，認定B2B行為是不正當的。72. § 4 No 4 UWG (‘gezielte Behinderung’); this doctrinal option was first brought to my attention by an oral statement of Matthias Leistner.作為這樣一種利益，根據反不正當競爭法的基本原理，人們可以突出整個市場的運行和競爭。其優勢尤其在于，在反壟斷法意義上沒有主導地位的情況下，市場失靈可以得到補救。73. On the antitrust framework for data access, cf Josef Drexl, ‘Designing Competitive Markets for Industrial Data’ (2017) 8 JIPITEC 257, 280 ff; under a UCL approach, it is also conceivable to draw on the antitrust criteria the CJEU has established in its ‘essential facility’ doctrine (cf Case C-418/01 IMS Health ECLI:EU:C:2004:257) as a starting point and then develop them further, duly heeding differences and specificities; on FRAND principles as a potential role model for data access, cf Heiko Richter and Peter R Slowinski, ‘The Data Sharing Economy: On the Emergence of New Intermediaries’ (2019) 50 IIC 4; in any case, aligning the ‘fairness’ element of FRAND with a claim based on unfair competition law appears apt at least on the terminological surface.在獲取數據方面，市場失靈是多方面的，不僅限于濫用壟斷權力的情形。74. For an overview of potential market failures relating to data access, see Bertin Martens, ‘Data Access, Consumer Interests and Social Welfare: An Economic Perspective’ (2020) ＜https://papers.ssrn.com/sol3/papers. cfm?abstract_id?3605383＞accessed before 27 November 2020.這也應該放在源于德國反壟斷法（《反限制競爭法》第20條）“輸出”的“相對市場支配地位”概念背景下進行看待，盡管在歐洲層面上缺乏類似規定，但其在規范數字經濟方面具有相當大的潛在意義。75. cf Heike Schweitzer at the Consumer Law Days 2019 (n 48) 231; Drexl (n 68) 33, 36, 41.

在反壟斷中，其他司法管轄區考慮采用這樣的考量因素，似乎并不令人信服； 相反，一個有效的選擇是將它們解釋為反不正當競爭法中的體系混合現象。從實質上講，某種權力不對稱（但低于支配閾值）可能會（共同）決定干預的衡量標準。76. cf Martin Peitz and Heike Schweitzer, ‘Ein neuer europa¨ischer Ordnungsrahmen fu¨ r Datenma¨ rkte?’ [2018] NJW 275,280, encouraging the development of case groups of ‘data-related exclusionary conduct’ in B2B relationships beyond market dominance constellations.

其次，尤其是在消費者的訪問意愿方面，反不正當競爭法似乎是理想的體系空間，因為其B2C維度通常被歸類為“消費者保護法”領域。77. See Drexl (n 48) 237, arguing that the constellation resembles the rules of advertising, a traditional key realm of UCL;comprehensively, Drexl (n 68) 40 ff; see also Jo¨ rg Hoffmann and Begonia Gonzalez Otero, ‘Demystifying the Role of Data Interoperability in the Access and Sharing Debate’ (2020) Max Planck Institute for Innovation & Competition Research Paper No 20-16, 20 ＜https://papers.ssrn.com/sol3/papers.cfm? abstract_id?3705217＞ accessed before 27 November 2020.在某種程度上，訪問與可移轉性相對應，基于反不正當競爭法的可移轉性機制可與通用數據保護條例（GDPR）的第20條的行為范式結合，在數字消費者福利的共同愿景下進行理論化。關于授予此類訪問權的實質性標準，有人建議，為了最佳使用連接設備，必須使用某些數據，并將該聲明構造為“連接性聲明”，甚至超出了可移轉性。78. According to Drexl (n 48) 238, it appears ‘fair’ to grant data access to consumers who need such access in order to use their device in an economically sound manner; on consumer access needs in the IoT, cf also Drexl (n 68); Josef Drexl, ‘Data access and control in the era of connected devices’ (2019) ＜https://www.beuc.eu/publications/beuc-x-2018-121_ data_access_and_control_in_the_area_of_connected_devices.pdf＞ accessed before 27 November 2020.

（二）反不正當競爭法與“人工智能與知識產權”對話，第1部分：市場敏感型投資保護

雖然整個學界似乎均在討論人工智能及其輸出的傳統知識產權，尤其是版權和專利保護，79. cf Jyh-An Lee, Kung-Chung Liu and Reto M Hilty (eds), Artificial Intelligence and Intellectual Property (OUP 2021)(forthcoming); Ryan Abbott, ‘I Think Therefore I Invent’ (2016) 57 Boston College Law Review 1079; Ana Ramalho, ‘Will robots rule the (artistic) world?’ (2017) 21 Journal of Internet Law 15; Annemarie Bridy, ‘Coding Creativity’ (2012) 5 Stanford Technology Law Review 1.但立足反不正當競爭法對相關主題的保護卻鮮有受到學界關注。80. But see Tim W Dornis, ‘Artificial Creativity: Emergent Works and the Void in Current Copyright Doctrine’ (2020) 22 Yale Journal of Law and Technology 1, 25 ff; Tim W Dornis, ‘Der Schutz ku¨ nstlicher Kreativita¨ t im Immaterialgu¨ terrecht’ [2019]GRUR 1252, 1256 f; Daniel Gervais, ‘Exploring the Interfaces between Big Data and Intellectual Property Law’ (2019) 10 JIPITEC 3, 19 para 84; Drexl (n 73) 270 para 62.是填補這一空白的時候了。81. Trade secret protection as a hybrid regime between IP and UCL will be considered separately in section IX.3. below.

1.通過反不正當競爭法保護人工智能創新：實踐風險和理論視野

一個長期而有爭議的討論圍繞著在多大程度上可以根據反不正當競爭法同時或在既有的知識產權法之外授予對模仿無形主體的保護。這些理論的具體設計在歐盟成員國和國際上各不相同。82. In the Anglo-American sphere, they appear as ‘misappropriation doctrine’, which is to some extent comparable to continental European UCL approaches, but very narrowly construed, cf Tim W Dornis, ‘Artificial Creativity’ (n 80) 26 ff.除了反不正當競爭法特定情形（如欺騙模仿）之外，鑒于傳統理論禁止基于“道德”理由的“照搬式”或“寄生式”模仿，83. This ‘moral’ rhetoric still resonates in the French terminology of ‘parasitisme’.無論是市場效應，還是事實上不斷擴大的知識產權保護范圍，現代理論強調了反不正當競爭法作為靈活且對市場敏感的保護機制的潛力。84. See Hilty (n 5); Ansgar Ohly, ‘A Fairness-Based Approach to Economic Rights’ in Bernt Hugenholtz (ed), Copyright Reconstructed (Wolters Kluwer 2018) 83; Annette Kur, ‘What to Protect, and How? Unfair Competition, Intellectual Property,or Protection Sui Generis’ in Nari Lee and others (eds), Intellectual property, unfair competition and publicity: convergences and development (Edward Elgar 2014) 11, 27 f; Ansgar Ohly, ‘The Freedom of Imitation and Its Limits - A European Perspective’(2010) 41 IIC 506, 522.認識到這一區別是以下考慮的關鍵: 然而在實踐中，反不正當競爭法的現有形式——如法院所適用的，仍部分地沿用著舊的道德原則——造成了過度保護公共領域事項的危險，85. See on this danger Drexl (n 73) 270 para 63.現代的、市場敏感型的經濟觀點具有相當大的潛力。這種潛力可能體現在三個方面：首先，在一個抽象的法律理論闡釋中，它象征著一種為數據經濟量身定制的無形商品保護方法的總體監管范式，即靈活性；其次，在不確定情況下，作為引入新的權利的替代方案，從中提取且與應然法相關的考量因素；第三，考慮到人工智能可能會從根本上改變知識產權格局，并相應地重塑與反不正當競爭法的互動局面，這是一個經典而又錯綜復雜的闡釋，與知識產權實然法的補充保護功能相關。雖然這三個維度顯然緊密聯系，但下面的分析將把它們作為一個寬泛的三重結構予以建構起來。

2.靈活的經濟功能侵權評估

從現代意義上反不正當競爭法的總體法律理論特性開始，這些似乎使其成為人工智能創新監管的完美匹配。簡言之，反不正當競爭法的保護是基于行為，而非主體導向;86. cf Drexl (n 73) 278 para 112: ‘This however questions the very appropriateness of a property approach to regulating that economy. IP systems are largely based on the paradigm of protecting intangible assets, such as technologies in particular, that play a role as input in the production of physical goods. Such a paradigm does not seem to fit a world in which customers have to rely on real-time and accurate information as an input.’它是高度靈活的，而不是依賴于標準化、預先確定的標準；它對福利經濟觀點很敏感，即在必要的程度內補救市場失靈——無論這種市場失靈是由知識產權領域的過度保護，還是保護不足造成的。不利的一面是缺乏法律確定性，經濟學知識的眾多不足之處以及實際應用的復雜性。87. cf Rupprecht Podszun, ‘Der ,,more economic approach“im Lauterkeitsrecht’ [2009] WRP 509, 517.在靈活性特點中值得注意的是，反不正當競爭法未有預先確定的條款：因此，從理論上講，它具有持續投資攤銷所需時間的潛力，88. See Markus Deck, ‘§ 17 Wettbewerblicher Nachahmungsschutz (§ 4 Nr. 3 UWG)’ in Gordian Hasselblatt (ed), MAH Gewerblicher Rechtsschutz (5th edn, CH Beck 2017) para 164, on the duration of protection for traditional computer programs under UCL.而正式保護期限與實際保護需要之間的差距長期以來一直被認為是知識產權法的一個危及福利的問題。89. cf Reto M Hilty and Thomas Jaeger, ‘Gesamtanalyse und Erkenntnisse’ in Reto M Hilty and Thomas Jaeger (eds),Europa¨isches Immaterialgu¨ terrecht - Funktionen und Perspektiven (Springer 2018) 665, 675.這在人工智能背景下變得更加重要，人工智能的特征具有十分動態的生產周期，難以與抽象的保護條款保持一致。90. See Hilty, Hoffmann and Scheuerer (n 37) 20; Drexl (n 73) 278 para 112: ‘In an environment where it is key to capture the moment and where being late leads to wrong decisions, asking the question of how long data should be protected will simply miss the needs of this economy.’此外，保護可以根據特定的行業需要進行調整，以因應人工智能行業特定的變化。91. In a way, contemplating sector-specific protection constitutes the mirror image of the current debate on sector-specific data access regimes.

關于行為依賴的特征，人工智能和物聯網領域的一個常見問題是難以界定92. This relates in particular to the dynamism of subject matter such as self-learning or ‘evolutionary’ algorithms.和定位93. This is reflected in the prominent debate on who should own the rights in ‘AI-generated’ output; on a more visionary note,a general blurring of ‘actors’ within global informational networks has been diagnosed, with the proposal of responding by ultimately holding ‘conduct itself’ liable, cf Gunther Teubner, ‘Digitale Rechtssubjekte?’(2018) 218 AcP 155, 202.保護對象。在這種可疑情形下，它似乎是一種可行的“解決方法”，而不是關注行為的福利效應，從而將問題從技術領域轉移到經濟領域。94. Comparable proposals have been made as to the reconstruction of copyright law: namely, instead of technically looking at ‘reproductions’, undertaking a ‘principle-based assessment’ inspired by modern trademark infringement doctrine, cf Taina Pihlajarinne, ‘Should We Bury the Concept of Reproduction - Towards Principle-Based Assessment in Copyright Law?’ (2017)48 IIC 953.此外，反不正當競爭法的經濟功能屬性似乎特別適合保護“人工智能生成的”無形物。法學界爭論中所關注的問題案例的特點是缺乏顯著的人力努力或指導。95. For a critical assessment of the technological state of the art vis-a` -vis scholarly ‘autonomy’ assumptions, see, however,Daria Kim, ‘AIGenerated Inventions: Time to Get the Record Straight?’ [2020] GRUR International 443.因此，反不正當競爭法的市場焦點似乎是一個特別合適的監管選擇。保護人類“創造者”（從廣義上理解，不限于版權）的人格和利益一直是授予知識產權的一個關鍵理由。96. See Hilty, Hoffmann and Scheuerer (n 37) 4 ff.然而，在沒有人的情況下，他們的利益必須在平衡工作中得到重視，“發明家”被“投資者”所取代，97. See Herbert Zech, ‘Artificial Intelligence: Impact of Current Developments in IT on Intellectual Property’ [2019] GRUR Int 1145, 1147: ‘Ultimately, AI generated innovations will only be protected or protectable by investment protection rights. The inventor (author) will be replaced by an investor using AI.’基于反不正當競爭法理論上合理的理由，對其采取“更多”而非“純粹經濟的方法”似乎是一個適當的框架。98. Robert Yu, ‘The Machine Author’ (2017) 165 University of Pennsylvania Law Review 1245, 1266 ff suggests using the ‘hot news misappropriation doctrine’, a ‘quasi-property-fairness-standard’, for handling AI creations.在每種情況下，均必須調查誰進行了相關投資，以及他們的補償是否因搭便車而受到威脅。從法律理論角度來看，一方面可以堅持區分歐洲大陸版權傳統中以人類為中心建立起的“經典”知識產權法，另一方面是人工智能的純粹經濟市場機制。99. Of course, this goes with the caveat that the ‘romantic’, anthropocentric understanding of IP has to a certain extent been overridden by industry-determined market realities, see Hilty, Hoffmann and Scheuerer (n 37) 27.

3.在市場失靈的不確定情況下替代新的知識產權

談及反不正當競爭法的應然法的考量因素時，反不正當競爭法傳統上被視為“標兵職能”，這意味著在相關理論最終成為完整的知識產權之前，可以基于反不正當競爭法的理由給予保護。100. On the ‘pacesetter function’ of UCL vis-a` -vis introducing new intellectual property rights, see Zech (n 10) 161 f; Ohly (n 84) 522 f; Kur (n 84) calls UCL an ‘incubator’ for new IP rights; emphasising the ‘interim’ character of a UCL solution in the AI context, Dornis, ‘Artificial Creativity’ (n 80) 44; Dornis, ‘Der Schutz ku¨ nstlicher Kreativita¨ t im Immaterialgu¨ terrecht’ (n 80)1252.在考慮潛在的新保護機制時，尤其是對于計算機生成的“作品”，以及數據或ML模型，應該記住這一屬性。101. cf Ce′line Castets-Renard, ‘The Intersection between AI and IP: Conflict or Complementarity?’ (2020) 51 IIC 141,142: ‘(...) the lawmaker may be led to consider that a sui generis system of IP rights for AIgenerated inventions should be raised to adjust innovation incentives for AI’; in favour of new IP regimes, Dornis, ‘Der Schutz ku¨ nstlicher Kreativita¨ t im Immaterialgu¨ terrecht’ (n 80) 1257 and 1264.只要不清楚引入此類權利是否有經濟需要，即是否存在需要補救的市場失靈，102. Outlining the context-dependency of market failure regarding AI outputs as opposed to not identifying market failure regarding AI tools, Hilty, Hoffmann and Scheuerer (n 37) 15 ff; considering market failure possible regarding training data,Philipp Hacker, ‘Immaterialgu¨ terrechtlicher Schutz von KI-Trainingsdaten’ [2020] GRUR 1025, 1033; assuming an economic need for protection, Dornis, ‘Der Schutz ku¨ nstlicher Kreativita¨ t im Immaterialgu¨ terrecht’ (n 80) 1264; yet all authors acknowledge the lack of clear empirical evidence. Absent such evidence, the whole market failure standard ultimately comes down to an allocation of the burden of proof or burden of justification, with the option of either the status quo or the freedom principle as a starting point.相反，人們可以全面監控事物的發展，收集經濟證據和見解，靈活地根據反不正當競爭法的理據給予保護，并在持續地保護需求實現后，將相關領域中確定的規則法典化。103. Critical on the introduction of new IP rights for trained AI, Zech (n 97) 1146: ‘Any reaction of IP law beyond jurisprudence and interpretative guidance has to be handled with care. New investment protection rights should only be introduced if otherwise a clear market failure is to be expected. In the area of artificial intelligence, this seems not to be the case’; on the sufficiency of(inter alia) UCL with regard to protection of AI data, cf also Peter R Slowinski, ‘Rethinking Software Protection’ (2020) Max Planck Institute for Innovation & Competition Research Paper No 20-17, 18 ＜https://papers.ssrn.com/sol3/papers.cfm?abstract_id?3708110＞ accessed before 27 November 2020.當然，基于不確定背景下潛在的功能失調的市場干預經濟成本，104. Dysfunctional effects of IP in the data economy are especially identified as regards the database sui generis right, see Matthias Leistner, ‘The Existing European IP Rights System and the Data Economy - An Overview With Particular Focus on Data Access and Portability’ (2020) 13 ff ＜https://papers.ssrn.com/sol3/papers.cfm?abstract_id?3625712＞ accessed before 27 November 2020.必須與在沒有明確定義權利105. Highlighting the problem of legal uncertainty when relying on UCL protection, Hacker (n 102) 1032; criticising UCL as‘rather shaky ground for the protection of industrial data’, Andreas Wiebe, ‘Protection of industrial data - a new property right for the digital economy?’ [2016] GRUR Int 877, 879; considering UCL ‘patchy at best’, Dornis, ‘Artificial Creativity’ (n 80) 59.的情況下的法律不確定性成本，以及與B2B領域的反不正當競爭法缺乏協調相關的成本進行權衡。106. Dornis, ‘Der Schutz ku¨ nstlicher Kreativita¨ t im Immaterialgu¨ terrecht‘(n 80) 1260; Wiebe (n 105) 879.

4.需要根據數字經濟的需要對理論要求進行微調

最后，關于反不正當競爭法保護的具體應用，應強調微調評估理論要求的必要性。法律秩序中常見的具體標準可分為兩類：首先，從歷史和體系角度來看，它們是與反不正當競爭法的其他傳統范式相一致的“特定不正當性”規范，尤其是對來源的混淆，即市場透明度問題，或者與違反商業秘密有關的知識收集；107. In Germany, for example, such criteria are codified in an explicit norm of the German Act against Unfair Competition (§4 No 3 UWG), while there is a long-standing discussion on under which circumstances further protection can be granted on the grounds of the general clause (cf Ansgar Ohly, ‘Hartplatzhelden.de oder: Wohin mit dem unmittelbaren Leistungsschutz?’ [2010]GRUR 487); Dornis, ‘Artificial Creativity’ (n 80) 27 considers ‘deceptive and goodwill-appropriating conduct’ as lying at the heart of misappropriation prevention in European and civil-law UCL.其次，它們可以作為知識產權保護閾值的功能等價物，如德國理論要求標的物呈現“競爭屬性”。這一標準一直存在疑問，而且，由于它依賴于視覺范式，其適用性也受到了數字化背景的新挑戰。108. cf Maximilian Becker, ‘Lauterkeitsrechtlicher Leistungsschutz fu¨ r Daten‘ [2017] GRUR 346, 347 f.然而，在法律方法的范圍內，此類標準通常可以由法院和學界靈活制定，因此，應根據數字經濟的需要和特點制定相應的標準。109. On the difficulties of applying ‘competitive originality’ to non-visual contexts, see Maximilian Becker, ‘§ 64 Lauterkeitsrechtlicher Leistungsschutz fu¨ r Daten’ in Wolfgang Gloy, Michael Loschelder and Rolf Danckwerts (eds),Handbuch des Wettbewerbsrechts (5th edn, CH Beck 2019) 47 ff.最終，他們的目標必須是在更具體的抽象層面上，為法院進行市場失靈評估提供指導。在此情形下，如果將基于反不正當競爭法的數據訪問機制的概念（見上文第IX.1節）與數據保護機制相結合，則綜合的反不正當競爭法方式具有逐步促進在訪問和保護之間找到廣泛尋求的最佳平衡的潛力。反不正當競爭法可以為從頭開始考慮的全新方法提供有利空間。一個具體的相關領域似乎正在推動特殊數據庫保護權的革新：要求進行這種改革的呼聲愈演愈烈，110. cf Leistner (n 104) 17 f; Drexl (n 68) 45.其中包括在形成一個新的和適當的機制時，需要將數據保護和數據訪問結合起來看待。

（三）反不正當競爭法與“人工智能與知識產權”對話，第2部分：人工智能組件的應用

當將這些考慮因素具體應用于人工智能時，它似乎傾向于按照機器學習過程的步驟來構建評估，即訓練數據、學習過程和輸出。111. This structure is inspired by Drexl and others (n 3).對這些現象的市場失靈的實質性評估超出了本文的范圍。112. For some literature cf n 102.相反，本文旨在闡明一些抽象的理論范式，以應對潛在的市場失靈。

1.訓練數據

從訓練數據開始，將反不正當競爭法保護應用于數據113. On the basic premise of what ‘data’ actually is, cf Zech (n 10) 32 f.總體上已經討論了很長一段時間，特別是作為一種替代方法或者反對在數據中引入新產權的論點。114. cf Josef Drexl and others, ‘Data Ownership and Access to Data - Position Statement of the Max Planck Institute for Innovation and Competition of 16 August 2016 on the Current European Debate’ (2016) Max Planck Institute for Innovation& Competition Research Paper No 16-10, para 18 ＜https://papers.ssrn.com/sol3/papers.cfm?ab stract_id?2833165＞ accessed before 27 November 2020; Drexl (n 73) 270; it is worth noting that although a purely economic perspective on the data property issue has in general rightfully been criticised as too short-sighted (cf Amstutz (n 53) 441), in the concrete doctrinal context of UCL as a competition-oriented regime, the standard must be an economic one; on data as subject matter of UCL protection, see generally Becker (n 108) 346; theoretically open towards applying the UCL general clause to data, Ansgar Ohly, ‘Anmerkung zu BGH, Unlauteres Verhalten als Voraussetzung fu¨ r wettbewerbsrechtlichen Nachahmungsschutz - Segmentstruktur’ [2017]GRUR 79, 92; Rupprecht Podszun, ‘§ 3 UWG’ in Henning Harte-Bavendamm and Frauke Henning-Bodewig (eds), UWG (4th edn, CH Beck 2016) para 178.如此一來，反不正當競爭法還可以構成一種防止數據的特定子現象（即人工智能訓練數據）被盜用的保護手段，這意味著防止通過使用與競爭對手相同的訓練數據創建另一個人工智能模型。115. cf Hacker (n 102) 1031, himself critical as to the sufficiency of such an approach.時間動態性的特點促使學者們對進行了數據的比較，并相應地考慮了在反不正當競爭法下對時尚的動態法律保護。假設兩者均具有很高的價值，但是短暫的，因此至少注冊的知識產權似乎不適合對其進行最佳保護。116. On protection regimes for fashion, see Kal Raustiala and Christopher Sprigman, ‘The Piracy Paradox: Innovation and Intellectual Property in Fashion Design’ (2006) 92 Virginia Law Review 1687, 1692; a further role model could be financial information for stock market transactions, cf Gervais (n 80) 9 para 30; for applying old misappropriation doctrines to new data contexts, cf also Victoria Ekstrand and Christopher Roush, ‘From “Hot News” to “Hot Data”: The Rise of FinTech, the Ownership of Big Data, and the Future of the Hot News Doctrine’ (2017) 35 Cardozo Arts & Entertainment LJ 303.無論人們認為這些相似是否令人信服，在任何（和每種）情況下，都必須考慮人工智能培訓數據的特定經濟特質，尤其是生成或獲取這些數據所需的投資。117. On the need to differentiate between personal data, industrial raw data and AI training data, cf Hacker (n 102) 1025.至于評估進一步取決于上述相應法律秩序的理論要求，118. According to Hacker (n 102) 1031, an example for UCL-specific ‘unfair conduct’ would be that employees take training data with them when changing their workplace to a competitor.特別是數據是否表現出“競爭性原創性”存在爭議，并且假設在大多數情況下它們沒有。119. Dismissive, Christoph Zieger and Nikolas Smirra, ‘Fallstricke bei Big Data-Anwendungen’ [2013] MMR 418, 421; very critical also Hacker (n 102) 1032.

2.算法和模型

關于人工智能算法的保護，必須在技術上和法律上加以區分：在訓練模型的基礎上，優化算法基本上由經典軟件組成，它們某種程度上是用計算機代碼編寫的，120. cf Zech (n 97) 1146.而此類算法永遠不受知識產權保護。因此，它們不僅在著作權法和專利法下的待遇與經典軟件相同，121. ibid.而且也適用反不正當競爭法對軟件保護的一般范式。122. In this context, it should be noted that the correct doctrinal realm for locating software protection has always been debated and the introduction of a ‘sui generis’ right been discussed as an alternative, cf Reto M Hilty and Christophe Geiger, ‘Patenting Software? A Judicial and SocioEconomic Analysis’ (2005) 36 IIC 615, 643 f; Slowinski (n 103) 9 ff; thus, when rethinking the respective paradigms against the backdrop of AI, UCL appears a valuable option to be taken into consideration in discourses on the apt systematic location of protection.在這方面，值得一提的是，上述關于計算機程序的流行觀點也是如此。123. On the comparability of fashion and (traditional) computer programs under UCL, see Deck (n 88) para 163.這些通常可以受到反不正當競爭法的保護，124. As with data, the established requirements of the doctrinal acquis, at least under German law, may pose some problems.As far as ‘competitive originality’ is understood as an indication of origin, it is doubtful to what extent computer programs in general and AI models in particular can be considered to display such character. Also, the traditional visual element associated with this notion hardly fits computer programs, cf Deck (n 88) para 159.然而必須注意著作權法和專利法關于它們（非）保護范圍的結論性決定，鑒于對這些現象特定的知識產權保護，反不正當競爭法的現實相關性幾乎缺失。這個案例更為復雜，比如訓練有素的人工智能模型，即實際的人工智能工具：這些模型，包括它們所構成的“權重”，是否或在多大程度上受版權法和專利法的保護存在爭議，125. cf Slowinski (n 103) 16 ff; Bego～na Gonzalez Otero, ‘Machine Learning Models Under the Copyright Microscope: Is EU Copyright Fit for Purpose?’ (2020) Max Planck Institute for Innovation & Competition Research Paper No 21-02 ＜https://ssrn.com/abstract? 3749233＞ accessed before 31 December 2020; Patrick Ehinger and Oliver Stiemerling, ‘Die urheberrechtliche Schutzfa¨higkeit von Ku¨ nstlicher Intelligenz am Beispiel von Neuronalen Netzen’ [2018] CR 2018 761.尤其是它們的動態、不斷變化的性質可能會改變傳統的知識產權范式。126. cf Zech (n 97) 1146, pointing at the possibility of UCL protection for trained AI.因此，一方面，與優化算法相比，可以認為反不正當競爭法的相關性更大，因為其基于行為的靈活性，將反不正當競爭法應用于訓練有素的ML模型可以靈活地解決這些不確定環境下的市場失靈。 另一方面，如果模型不受知識產權軟件保護，則此決定通常不應被反不正當競爭法保護所規避或者推翻。

3.生成物

對于人工智能生成的產物，遵循知識產權法的系統決策必須再次成為適用反不正當競爭法的關鍵指引。如前所述，它的潛在相關性尤其體現于由于缺乏人類作者、發明家或設計師而無法為人工智能生成物提供知識產權保護的情形下。127. cf Zech (n 97) 1146; Dornis, ‘Der Schutz ku¨ nstlicher Kreativita¨ t im Immaterialgu¨ terrecht’ (n 80) 1256 f.然而，無論這種缺失是系統性的故意還是偶然的仍然模棱兩可：“人工智能生成物”在相關法律頒布時根本無法想象，這一論點可以轉向一個或者另一個方向。128. Dornis, ‘Der Schutz ku¨ nstlicher Kreativita¨ t im Immaterialgu¨ terrecht’ (n 80) 1257 identifies a ‘gap’ of IP law regarding this issue.無論如何，與人工智能生成的無形物品相關的兩項進展需要密切監測：第一，對市場結構的經驗經濟學洞察，以確定市場失靈或其缺失；第二，關于缺乏人參與可能對“公共領域”的知識產權范式產生的影響的法律理論討論。129. cf Mauritz Kop, ‘AI & Intellectual Property: Towards an Articulated Public Domain’ (2020) 29 Texas Intellectual Property Law Journal ＜https://papers.ssrn.com/sol3/papers.cfm?abstract_id?3409715＞ accessed before 27 November 2020.

（四）商業秘密保護的反不正當競爭法維度

最后，商業秘密保護的反不正當競爭法維度可以作為監管人工智能經濟的法律理論基石。在歐盟，商業秘密保護同一時期已被編纂為一項獨立的法律。130. Directive (EU) 2016/943 on the protection of undisclosed knowhow and business information (trade secrets) against their unlawful acquisition, use and disclosure.然而，它仍然植根于并援引反不正當競爭法，特別是當它依賴（不）正當標準作為侵權法的附屬一般條款時。131. cf Drexl (n 78) 97: ‘Here, the Directive integrates EU trade secrets protection into a broader unfair competition law framework.’商業秘密保護是人工智能/知識產權保護領域的一個重要組成部分。132. Josef Drexl and others, ‘Comments of the Max Planck Institute for Innovation and Competition of 11 February 2020 on the Draft Issues Paper of the World Intellectual Property Organization on Intellectual Property Policy and Artificial Intelligence’(2020) 9 ＜https://pure.mpg. de/rest/items/item_3193085_2/component/file_3193086/content＞ accessed before 27 November 2020.數據、算法、模型和生成物均可作為商業秘密進行保護。133. cf Tanya Aplin, ‘Trading Data in the Digital Economy: Trade Secrets Perspective’ in Sebastian Lohsse, Reiner Schulze and Dirk Staudenmayer (eds), Trading Data in the Digital Economy: Legal Concepts and Tools (Nomos 2017); specifically on trade secret protection for AI training data, Hacker (n 102) 1032.盡管存在某些福利主義的模糊性，134. The effects of trade secret protection on AI innovation are ambivalent insofar as on the one hand, the regime provides some extent of exclusivity, thereby protecting investments and safeguarding innovation incentives, while on the other hand, it also creates obstacles vis-a` -vis third parties that want to use e.g. certain data to train their ML models.但歐洲商業秘密制度作為一個平衡且充分的制度在排他性和獲取性之間實現最佳權衡而被廣泛贊譽。135. Leistner (n 104) 18 ff.這主要歸功于它采取了受到反不正當競爭法啟發的、靈活的、基于行為的方法。136. Drexl (n 73) 269 para 56: ‘(...) such further limited protection can be considered as better suited to serve the purposes of the data economy, by focussing on the particular way in which a third party has specifically acquired access to the data instead of granting exclusive protection against the use of data’.該制度不是一個成熟的財產角度，137. ibid 291, para 182: ‘Rather than recognising exclusive control over any use of protected information, as would be typical for intellectualproperty regimes, EU trade secrets law implements a tort law approach that bans specific conduct related to the acquisition, dissemination and use of trade secrets that can be considered unfair.’而是作為知識產權和反不正當競爭法的理論混合體構建的，將彼此的優點結合起來。138. On the advantages of legal hybrids, see Ohly (n 84) 86 ff.根據這些優點，商業秘密指令可被視為上述反不正當競爭法方法的整體法律理論特征的具體化。不僅是為了法律的一致性，商業秘密保護和反不正當競爭法不應被視為兩個獨立的領域，而是應當（仍然）以相互的觀點理解和解釋。畢竟，TS指令對反不正當競爭法標準的明確依賴性，也有可能使反不正當競爭法作為一個利益和重要領域重新煥發活力，并在其B2B維度上重新推動歐洲的協調對話。

十、結論與展望

事實證明，在日漸由人工智能決定的市場秩序方面，反不正當競爭法將發揮切實可行的作用。它可以在許多方面促進中央監管模式的落實，旨在為了社會利益最佳的利用這一新技術。因此，反不正當競爭法能夠而且不應該只是被動地因應和調整其既有標準，而是應憑借其理論靈活性，積極參與制定應對人工智能所提出的多種挑戰所需的新標準。與此同時，人工智能所引發的法律問題為進一步反思和完善反不正當競爭法的本質和核心提供了契機，反不正當競爭法是一個仍未有得到充分理解的法律體系。由于其特有的靈活性，反不正當競爭法對社會、經濟和技術變革表現出高度的依賴性和回應性。這些變化目前和今后一段時間均是由人工智能推動的。反不正當競爭法可能會以這種方式被重塑與改進，成為調整數字經濟、競爭秩序和社會的真正基石。

致謝

感謝弗諾克·亨寧·博德維希教授、約爾格·霍夫曼和克勞斯·維德教授提供的有益意見。