生成式AI的法律风险及应对Legal Risks and Responses to Generative AI
何煊
摘要(Abstract)：
随着生成式AI应用在多个领域的兴起，在为经济发展、科技进步注入活力的同时，也存在侵害用户权益、损害公共利益的可能。为消弭技术可能带来的多维度风险，监管需求与日俱增。生成式AI模型与AI系统既存在区别又彼此联系，共同在特定场景下发挥作用，构成对生成式AI规制的逻辑起点。以《AI法案》为代表的欧盟立法具有一定的前瞻性，却也在如风险分级、生成内容审核等多方面存在问题，需要持续完善。生成式AI非普通产品，考虑其动态性与场景化的特征，应当采取具有针对性的规制措施。我国立法可适当吸取有益经验，在价值链主体合作机制构建与平台生成内容审核方面予以创新，提高生成式AI规制的有效性。
关键词(KeyWords)： 生成式AI;法律风险;《AI法案》;价值链;平台;内容审核
基金项目(Foundation):
作者(Author): 何煊
DOI: 10.14084/j.cnki.cn62-1185/c.2023.04.008
参考文献(References)：
- [1]RADFORD A,WU J,CHILD R,et al.Language Models are Unsupervised Multitask Learners[J].OpenAI blog,2019,1(8). [2]于文轩，马亮，王佃利，等.“新一代人工智能技术ChatGPT的应用与规制”笔谈[J].广西师范大学学报(哲学社会科学版),2023,59(2). [3]BALLARDINI R M,HE K,ROOS T.AI-generated Content:Authorship and Inventorship in the Age of Artificial Intelligence[C].Online Distribution of Content in the EU,2019. [4]CRESWELL A,WHITE T,DUMOULIN V,et al.Generative Adversarial Networks:An Overview[J].IEEE Signal Processing Magazine,2018,35(1). [5]蒲清平，向往.生成式人工智能——ChatGPT的变革影响、风险挑战及应对策略[J/OL].重庆大学学报(社会科学版)2023(3)[2023-04-20].http://kns.cnki.net/kcms/detail/50.1023.C.20230412.1004.002.html. [6]徐伟.论生成式人工智能服务提供者的法律地位及其责任——以ChatGPT为例[J/OL].法律科学(西北政法大学学报),2023(4)[2023-04-13].https://doi.org/10.16290/j.cnki.1674-5205.2023.04.013. [7]JAKHAR D,KAUR I.Artificial Intelligence,Machine Learning and Deep Learning:Definitions and Differences[J].Clinical and Experimental Dermatology,2020,45(1). [8]GOODFELLOW I,BENGIO Y,COURVILLE A.Deep Learning[M].MIT press,2016. [9]张凌寒.深度合成治理的逻辑更新与体系迭代——ChatGPT等生成型人工智能治理的中国路径[J].法律科学(西北政法大学学报),2023,41(3). [10]DE ANGELIS S F.Artificial Intelligence:How Algorithms Make Systems Smart[EB/OL].(2014-09-19)[2023-03-19].https://www.wired.com/insights/2014/09/artificial-intelligence-algorithms-2/. [11]GUTIERREZ C I,Aguirre A,UUK R,et al.A Proposal for a Definition of General Purpose Artificial Intelligence Systems[EB/OL].(2022-05-08)[2023-03-08].https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4238951. [12]RUSSELL S J.Artificial Intelligence a Modern Approach[M].Pearson Education,Inc.,2010. [13]OUYANG L,WU J,JIANG X,et al.Training Language Models to Follow Instructions with Human Feedback[J].Advances in Neural Information Processing Systems,2022,35. [14]THORP H H.ChatGPT is Fun,But Not an Author[J].Science,2023,379(6630). [15]SALAKHUTDINOV R.Learning Deep Generative Models[J].Annual Review of Statistics and Its Application,2015(2). [16]RADFORD A,NARASIMHAN K,SALIMANS T,et al.Improving Language Understanding by Generative Pre-training[EB/OL].(2023-05-02)[2023-05-03].https://www.mikecaptain.com/resources/pdf/GPT-1.pdf. [17]LOPES R,BOYEAU P,YOSEF N,et al.Decision-making with Auto-encoding Variational Bayes[J].Advances in Neural Information Processing Systems,2020,33. [18]毕文轩.生成式人工智能的风险规制困境及其化解：以ChatGPT的规制为视角[J].比较法研究，2023(3). [19]刘艳红.生成式人工智能的三大安全风险及法律规制——以ChatGPT为例[J/OL].东方法学，2023(4)[2023-06-06].https://doi.org/10.19404/j.cnki.dffx.20230606.003. [20]唐林垚.具身伦理下ChatGPT的法律规制及中国路径[J].东方法学，2023(3). [21]KOP M.EU Artificial Intelligence Act:The European Approach to AI[EB/OL].(2021-10-01)[2023-03-09].https://law.stanford.edu/wp-content/uploads/2021/09/2021-09-28-EU-Artificial-Intelligence-Act-The-European-Approach-to-AI.pdf. [22]HORNE B D,ADALI S.This Just In:Fake News Packs a Lot in Title,Uses Simpler,Repetitive Content in Text Body,More Similar to Satire than Real News[J].Proceedings of the international AI conference on web and social media,2017,11(1). [23]陈全真.生成式人工智能与平台权力的再中心化[J].东方法学，2023(3). [24]宋华健.反思与重塑：个人信息算法自动化决策的规制逻辑[J].西北民族大学学报(哲学社会科学版),2021(6). [25]梁上上.利益衡量的界碑[J].政法论坛，2006(5). [26]EDWARDS L.Regulating AI in Europe:Four Problems and Four Solutions[EB/OL].(2022-03-31)[2023-03-06].https://www.adalovelaceinstitute.org/report/regulating-ai-in-europe/. [27]CASTELLS M.The Internet galaxy:Reflections on the Internet,Business,and Society[M].Oxford University Press,USA,2002. [28]HELBERGER N,DIAKOPOULOS N.ChatGPT and the AI Act[EB/OL].(2023-02-16)[2023-03-15].https://policyreview.info/pdf/policyreview-2023-1-1682.pdf. [29]MCKOWN J R.Discovery of Trade Secrets[J].Santa Clara Computer & High Tech.LJ,1994(10). [30]Verbraucherleitbilder:interdisziplin?re und europ?ische Perspektiven[M].Walter de Gruyter GmbH & Co KG,2015. [31]DEEKS A.The Judicial Demand for Explainable Artificial Intelligence[J].Columbia Law Review,2019,119(7). [32]WACHTER S,MITTELSTADT B,FLORIDI L.Why a Right to Explanation of Automated Decision-making does not Exist in the General Data Protection Regulation[J].International Data Privacy Law,2017,7(2).
- (1)参见《互联网信息服务深度合成管理规定》第23条。 (2)See Article 3 AI Act. (3)See European Parliament:General Product Safety Directive(2021). (4)该文详细介绍了加密水印的定义及作用机理,其被认为能够有效维护数据、隐私与个人信息安全。“How The ChatGPT Watermark Works And Why It Could Be Defeated”Search Engine Journal,2023-04-28,https://www.searchenginejournal.com/chatgpt-watermark/475366/. (5)See Article 3(g)(iii) DSA. (6)See Annexes II Section A.No.12,13 and III No.3-5 AI Act. (7)See Article 2(8) AI Act. (8)See Article (5) AI Act:Providers of general purpose AI systems shall cooperate with and provide the necessary information to other providers intending to put into service or place such systems on the Union market as high-risk AI systems or as components of high-risk AI systems,with a view to enabling the latter to comply with their obligations under this Regulation. (9)See Article 70(1) AI Act. (10)See Article 11,Annex IV AI Act. (11)See Article 11,DSA.