人工智能治理聯盟：生成式人工智能治理塑造集體全球未來（英文版）（21頁）.pdf

《人工智能治理聯盟：生成式人工智能治理塑造集體全球未來（英文版）（21頁）.pdf》由會員分享，可在線閱讀，更多相關《人工智能治理聯盟：生成式人工智能治理塑造集體全球未來（英文版）（21頁）.pdf（21頁珍藏版）》請在三個皮匠報告上搜索。

1、I N C O L L A B O R A T I O N W I T H A C C E N T U R EGenerative AI Governance:Shaping a Collective Global Future3/3AI Governance AllianceBriefing Paper Series 2024ContentsImages:Getty Images,MidJourney 2024 World Economic Forum.All rights reserved.No part of this publication may be reproduced or t

2、ransmitted in any form or by any means,including photocopying and recording,or by any information storage and retrieval system.Disclaimer This document is published by the World Economic Forum as a contribution to a project,insight area or interaction.The findings,interpretations and conclusions exp

3、ressed herein are a result of a collaborative process facilitated and endorsed by the World Economic Forum but whose results do not necessarily represent the views of the World Economic Forum,nor the entirety of its Members,Partners or other stakeholders.Executive summaryIntroduction1 Global develop

4、ments in AI governance1.1 Evolving AI governance tensions2 International cooperation and jurisdictional interoperability2.1 International coordination and collaboration2.2 Compatible AI standards2.3 Flexible regulatory mechanisms3 Enabling equitable access and inclusive global AI governance3.1 Struc

5、tural limitations and power imbalances3.2 Inclusion of the Global South in AI governanceConclusionContributorsEndnotes34568899101011121317Generative AI Governance:Shaping a Collective Global Future2Executive summaryThe global landscape for artificial intelligence(AI)governance is complex and rapidly

6、 evolving,given the speed and breadth of technological advancements,as well as social,economic and political influences.This paper examines various national governance responses to AI around the world and identifies two areas of comparison:1.Governance approach:AI governance may be focused on risk,r

7、ules,principles or outcomes;and whether or not a national AI strategy has been outlined.2.Regulatory instruments:AI governance may be based on existing regulations and authorities or on the development of new regulatory instruments.Lending to the complexity of AI governance,the arrival of generative

8、 AI raises several governance debates,two of which are highlighted in this paper:1.How to prioritize addressing current harms and potential risks of AI.2.How governance should consider AI technologies on a spectrum of open-to-closed access.International cooperation is critical for preventing a fract

9、uring of the global AI governance environment into non-interoperable spheres with prohibitive complexity and compliance costs.Promoting international cooperation and jurisdictional interoperability requires:International coordination:To ensure legitimacy for governance approaches,a multistakeholder

10、approach is needed that embraces perspectives from government,civil society,academia,industry and impacted communities and is grounded in collaborative assessments of the socioeconomic impacts of AI.Compatible standards:To prevent substantial divergence in standards,relevant national bodies should i

11、ncrease compatibility efforts and collaborate with international standardization programmes.For international standards to be widely adopted,they must reflect global participation and representation.Flexible regulatory mechanisms:To keep pace with AIs fast-evolving capabilities,investment in innovat

12、ion and governance frameworks should be agile and adaptable.Equitable access and inclusion of the Global South in all stages of AI development,deployment and governance is critical for innovation and for realizing the technologys socioeconomic benefits and mitigating harms globally.Access to AI:Acce

13、ss to AI innovations can empower jurisdictions to make progress on economic growth and development goals.Genuine access relies on overcoming structural inequalities that lead to power imbalances for the Global South,including in infrastructure,data,talent and governance.Inclusion in AI:To adequately

14、 address unique regional concerns and prevent a relegation of developing economies to mere endpoints in the AI value chain,there must be a reimagining of roles that ensure Global South actors can engage in AI innovation and governance.The findings of this briefing paper are intended to inform action

15、s by the different actors involved in AI governance and regulation.These findings will also serve as a basis for future work of the World Economic Forum and its AI Governance Alliance that will raise critical considerations for resilient governance and regulation,including international cooperation,

16、interoperability,access and inclusion.Shaping a prosperous and equitable global future with AI depends on international cooperation,jurisdictional interoperability and inclusive governance.Generative AI Governance:Shaping a Collective Global Future3IntroductionGenerative AI promises economic growth

17、and social benefits but also poses challenges.The rapid onset of generative artificial intelligence(AI)is promising socially and economically,1 including the potential to raise global gross domestic product(GDP)by 7%over a 10-year period.2 At the same time,a range of complex challenges has emerged,s

18、uch as the impact on employment,education and the environment,as well as the potential amplification of online harms.3 Additionally,there are increased demands for corporate transparency of AI systems4 and for clarity on data provenance and ownership.5 Governance authorities worldwide face the daunt

19、ing task of developing policies that harness the benefits of AI while establishing guardrails to mitigate its risks.Additionally,they are attempting to reconcile AI governance approaches with existing legal structures such as privacy and data protection,human rights,including rights of the child,int

20、ellectual property and online safety.Generative AI Governance:Shaping a Collective Global Future4Global developments in AI governance1The nascent and fragmented global AI governance landscape is further complicated by challenges posed by generative AI.The complex and fast-evolving AI governance land

21、scape is marked by diverse national responses:risk-based,rules-based,principles-based and outcomes-based,as delineated in Table 1.It is important to note the difficulty of neatly attributing singular approaches to individual jurisdictions,as elements of multiple approaches can complement each other

22、and are likely to be incorporated into hybrid responses.6Summary of AI governance approaches(not mutually exclusive)TABLE 1Risk-based Rules-based Principles-based Outcomes-based DefinitionFocuses on classifying and prioritizing risks in relation to the potential harm AI systems could causeLays out d

23、etailed and specific rules,standards and/or requirements for AI systemsSets out fundamental principles or guidelines for AI systems,leaving the interpretation and exact details of implementation to organizationsFocuses on achieving measurable AI-related outcomes without defining specific processes o

24、r actions that must be followed for complianceBenefits Tailored to application area Proportional to risk profile Flexible to changing risk levels Potential reduction of complexity Consistent enforcement possible Intended to foster innovation Adaptable to new developments Can encourage sharing of bes

25、t practices Can support efficiency Flexible to change Intended to foster innovation Compliance can be cost-effectiveChallenges Risk assessments can be complex May create barriers to market entry in high-risk areas Assessment and enforcement can be complex Rigidity can increase compliance costs May b

26、e unreliable to enforce Potential inconsistencies with interpretation of principles Unpredictable compliance and impractical enforcement Potential for abuse by bad actors Scope of measurable outcomes can be vague Potential for diffused accountability Limited control over process and transparencyExam

27、pleEU:Artificial Intelligence Act,2023(provisional agreement)China:Interim Measures for the Management of Generative AI Services,2023Canada:Voluntary Code of Conduct for Artificial Intelligence,2023Japan:Governance Guidelines for Implementation of AI Principles Ver.1.1,2022Generative AI Governance:S

28、haping a Collective Global Future5The recent provisional agreement reached on the EU AI Act represents the worlds first attempt at enacting comprehensive and binding AI regulation applicable to AI products and services within a risk-based and use case-driven structure.7 Other AI-specific regulatory

29、efforts are also under development in various jurisdictions,such as in Canada,8 Brazil,9 Chile10 and the Philippines.11 Meanwhile,the Indian government has weighed a non-regulatory approach,emphasizing the need to innovate,promote and adapt to the rapid advancement of AI technologies.12 In direct re

30、sponse to the rapid progress and widespread use of generative AI foundation models,China enacted regulations related to the use of generative AI.The EU AI Act also incorporates specific obligations for foundation models underpinning general-purpose AI(GPAI)systems.13Additional countries such as Sing

31、apore,14 Malaysia,15 Saudi Arabia,16 Japan,17 and Rwanda18 are responding to the transformative potential of AI by developing national polices19 that outline governance intentions and explore a range of regulatory instruments,ranging from hard laws and mandatory compliance rules to soft guidance and

32、 voluntary best practices.Lending to the intricacy of the governance landscape,regulatory responses are spread across a matrix of sector-specific considerations and cross-sectorial requirements.The recently issued US Executive Order on Safe,Secure,and Trustworthy Artificial Intelligence directs fede

33、ral agencies to develop new standards and includes sector-specific guidance driven by risk management.In addition to government regulatory efforts,there is a growing awareness of the importance of industry-responsible AI governance practices20 in safeguarding societal interests.For example,in respon

34、se to the US Executive Order the National Institute of Standards and Technology(NIST)has established the AI Safety Consortium,which intends to collaborate closely with industry,among other stakeholders,to inform risk management best practices.21The existence of a spectrum of AI governance approaches

35、 considers debates arising from new and amplified challenges22 introduced by the scale,power and design of generative AI technologies.Table 2 provides a snapshot of two prominent debates taking place with a sample of divergent positions regarding the nature of risks and access to AI models.Other eme

36、rging tensions include how generative AI will impact employment,23 its intersection with copyright protections,24 data transparency requirements,25 allocation of responsibility among actors within the generative AI life cycle26 and addressing misinformation and disinformation concerns amplified by g

37、enerative AI.27Many of these emerging tensions have their roots in data governance issues,28 such as privacy concerns,data protection,embedded biases,29 identity and security challenges from the use of data to train generative AI systems,and the resultant data created by generative AI systems.There

38、is a need to re-examine existing legal frameworks that provide legal assurance to the ownership of AI-generated digital identities.30Evolving AI governance tensions1.1Generative AI Governance:Shaping a Collective Global Future6Areas of debate in AI governance(non-exhaustive)TABLE 2Debate and context

39、Sample positionPolicy arguments forPolicy arguments againstPolicy focus on long-term existential risks31 vs present AI harms.32AI poses present harms and a spectrum of potential near-to long-term risks.Diverse positions exist regarding how to identify and prioritize the harms and risks from AI as we

40、ll as the timeframe over which risks should be considered.Advanced autonomous AI systems pose an existential threat to humanity.33 Without sufficient caution,humans could irreversibly lose control of autonomous AI systems.34 Starting with the biggest questions around existential risk supports the de

41、velopment of trustworthy AI and could prevent overregulation.35 Existential risks are speculative and uncertain.36 Can redirect the flow of valuable resources from scientifically studied present harms.37 Misdirects regulatory attention.38Effective regulation of AI needs grounded science that investi

42、gates present harms.39 In terms of urgency,there are immediate problems and emerging vulnerabilities with AI that disproportionately impact marginalized and vulnerable populations.Contending with known harms will address long-term hypothetical risks.40 Focus on known harms may lead to neglecting lon

43、g-term risks not well considered by traditional policy goals.Policy treatment of open-source vs closed-source AI.41Governance consideration is being given regarding where an AI technology may sit on a spectrum of open-to-closed access.42Open-source AI is critical to AI adoption and mitigating curren

44、t and future harms from AI systems.43 Increased access to AI and democratization of its capabilities.Spurs innovation and stimulates competition.Enables study of risks that can reduce bias and disparate performance for marginalized populations.Increased access exposes AI models to greater malicious

45、use and unintentional misuse.Difficulties in patching vulnerabilities can leave the AI system unsecured.44 Closed-source AI is necessary to protect against misuse of powerful AI technology.45 Protects commercial intellectual property.Safeguards against potentially harmful future capabilities.Identif

46、ied vulnerabilities can be fixed and safety features can be implemented.46 Concentration of power and knowledge within high-resource organizations.47 Increased dependency on a few foundation model providers with the risk of monopoly-related consequences.Generative AI Governance:Shaping a Collective

47、Global Future7International cooperation and jurisdictional interoperability2International cooperation to facilitate jurisdictional interoperability is vital to ensure global cohesion and trust in AI.International cooperation is critical to ensure societal trust in generative AI and to prevent a frac

48、turing of the global AI governance environment into non-interoperable spheres with prohibitive complexity and compliance costs.Facilitating jurisdictional interoperability requires international coordination,compatible standards and flexible regulatory mechanisms.For example,the US has taken the ini

49、tiative to enable cooperation with Europe through the US-EU Trade and Technology Council,while Chile,New Zealand and Singapore have signed a Digital Economy Partnership Agreement.Indicative of a growing consensus on the need for AI regulation,delegate nations at the 2023 UK AI Safety Summit signed t

50、he Bletchley Declaration with a commitment to establish a shared understanding of AI opportunities and risks.To ensure enduring legitimacy for governance proposals,global regulatory interoperability must adopt a multistakeholder approach that embraces a diversity of perspectives from government,civi

51、l society,academia,industry and impacted communities.Effective grounding of efforts in a comprehensive assessment of the socioeconomic impacts of AI and the efficacy of regulatory responses demands collaboration in identifying and prioritizing critical issues.Examples of international coordination e

52、fforts in drafting AI policy guidance include UNICEFs 2021 Policy guidance on AI for children and INTERPOLs 2023 Toolkit for Responsible AI Innovation in Law Enforcement developed in collaboration with the United Nations Interregional Crime and Justice Research Institute(UNICRI).Efforts like the Org

53、anisation for Economic Co-operation and Developments OECD.AI to map interoperability gaps between national governance frameworks48 are crucial to reducing conflicting regulatory requirements and establishing predictability and clarity for companies and people.At the intergovernmental level,coordinat

54、ion efforts to address international AI governance matters are currently under way at the Council of Europes Committee on AI,OECDs Working Party on Artificial Intelligence Governance,the African Union High-Level Panel on Emerging Technologies(APET),the Association of Southeast Asian Nations(ASEAN)wo

55、rkshops49 and the Guide on AI Governance and Ethics,50 the G751 and the G20,among others.52 In May 2023,G7 leaders published a report on the Hiroshima Process on Generative AI to study the rapidly evolving technology and help guide discussions on common policy priorities related to generative AI.53

56、Additionally,international efforts like the United Nations High-Level Advisory Body on AI and the World Economic Forums AI Governance Alliance are playing a critical role in coordinating multistakeholder dialogue and knowledge sharing to inform governance interoperability conversations.International

57、 coordination and collaboration2.1Generative AI Governance:Shaping a Collective Global Future8Governing bodies around the world are turning to standards as a method for governing AI.The British Standards Institution launched an AI Standards Hub aimed at helping AI organizations in the UK understand,

58、develop and benefit from international AI standards.The European Telecommunications Standards Institute(ETSI)and the European Committee for Electrotechnical Standardization(CENELEC)have published the European Standardization agenda that includes the adoption of external international standards alrea

59、dy available or under development,in part stimulated by the proposed EU AI Regulations framework for standards.In the US,NIST has developed an AI Risk Management Framework to support technical standards for trustworthy AI.54Despite criticisms regarding the instrumentalization of standards to shift r

60、egulatory powers from governments to private actors,55 they are increasingly recognized as an important tool in international trade,investment,competitive advantage and national values.There is concern that substantial divergences in approaches to setting AI standards threaten a further fragmentatio

61、n of the international AI governance landscape,lending to downstream social,economic and political implications internationally.International standardization programmes are being developed by the Joint Technical Committee of the International Organization for Standardization and the International El

62、ectrotechnical Commission(ISO/IEC JTC1/SC42)56 as well as by the Institute of Electrical and Electronic Engineers Standards Association(IEEE SA).For their part,the US,EU and China,have signalled commitments to undertake best efforts to align with internationally recognized standardization efforts.57

63、 Despite these signals,there is no guarantee that every country will follow these standards,especially if there is concern that their development has not been inclusive of local interests.Creating the capacity and space for broader participation in the standards-making process is thus needed.The fas

64、t-evolving capabilities of generative AI require investment in innovation and governance frameworks that are agile and adaptable.This includes ongoing assessment of opportunity and risk emanating from applied practice and feedback from those directly impacted by the technology.Flexible regulatory me

65、chanisms,beyond statutory instruments,are needed to account for societal implications and regulatory challenges that will emerge as generative AI technologies continue to advance and be adopted across various cultures and sectors.For example,Singapore,58 the United Arab Emirates,59 Brazil,60 the UK,

66、61 the EU,62 and Mauritius63 have pioneered“regulatory sandboxes”that allow organizations to test AI in a safe and controlled environment.Such policy innovations must be coupled with additional efforts to clarify regulatory intent and the associated requirements for compliance.For flexible mechanism

67、s to scale,supervisory authorities will need to consider how they provide industry participants confidence to participate and help establish agile best practice approaches while addressing the fear of regulatory capture through participation.Compatible AI standardsFlexible regulatory mechanisms2.22.

68、3 Creating the capacity and space for broader participation in the AI standards-making process is needed.Generative AI Governance:Shaping a Collective Global Future9Enabling equitable access and inclusive global AI governance3The Global Souths role in AI development and governance is critical to sha

69、ping a responsible future.The need for diversity and more equitably deployed generative AI systems is of significant global concern.Inclusive governance that consults with diverse stakeholders,including from developing countries,can help surface challenges,priorities and opportunities to make genera

70、tive AI technologies work better for everyone64 and address widening inequalities associated with the pre-existing digital divide.By ensuring the inclusion of underrepresented countries from Sub-Saharan Africa,the Caribbean and Latin America,the South Pacific,as well as some from Central and South A

71、sia(collectively referred to as the Global South)in international discussions on AI governance,a more diverse and equitable deployment of generative AI systems and compatibility of governance regimes can be achieved.The Global Souths priorities in areas such as healthcare,education or food security

72、oftenforce trade-offs,hampering investments in long-termdigital infrastructure.However,access to AI innovations can empower countries to make progress on economic growth and development goals65 where needs are greatest transforming health services,improving education quality,increasing agricultural

73、productivity,etc.to improve lives.66 Successfully deploying generative AI solutions at scale relies on overcoming several structural inequalities lending to power imbalances as detailed in Table 3.Structural limitations and power imbalances3.1Generative AI Governance:Shaping a Collective Global Futu

74、re10In addition to equitable access,inclusion of the Global South in all stages of the development and governance of AI is essential to prevent a reinforced power imbalance whereby developing economies are relegated to mere endpoints in the global generative AI value chain,either as extractive digit

75、al workers or as consumers of the technology.Though AI policy and governance frameworks are predominantly being developed in China,the EU and North America(46%),compared to 5.7%in Latin America and 2.4%in Africa,72 it is important to recognize the significant activities of different national bodies

76、such as Colombia,73 Brazil,74 Mauritius,75 Rwanda,76 Sierra Leone,77 Viet Nam78 and Indonesia,79 the recently introduced Digital Forum of Small States(FOSS)chaired by Singapore,as well as the emergence of AI research and industry ecosystems out of the Global South.The absence of historical and geopo

77、litical contexts of power and exploitation from dominant AI governance debates underscores the necessity for diverse voices and multistakeholder perspectives.The significant differences between some concerns of the Global South and those elevated within more dominant discourses of AI risks80 warrant

78、 a restructuring of AI governance processes,moving beyond current frameworks of inclusion.81 To adequately address regional concerns there must be a reimagining of roles that ensure Global South actors can engage in co-governance.Inclusion of the Global South in AI governance3.2Sources of global dis

79、parities and exclusion in generative AI(non-exhaustive)TABLE 3DimensionContextGovernance considerationsInfrastructure Access to compute,cloud providers and energy resourcesTraining generative AI systems,supporting experimentation and solution development and maintaining physical data centres67 requi

80、res extensive compute and cloud infrastructure that is financially and environmentally costly68 and results in high energy intensity.69The level of computing infrastructure required for research and development of generative AI models is primarily accessible to just a few industry laboratories with

81、sufficient funding.70 This puts at risk the participation of the vast majority in the development of these advanced models.Data Low resource languages and representation Generative AIs outputs inherently reflect the data and design of a models training.Current major generativeAI models are primarily

82、 developed in the US and China and trained on data from North America,Europe and China.Active inclusion of developing nations and diverse voices in generative AI development and governance is critical to ensure global inclusion in a future influenced by generative AI.Talent Access to education and t

83、echnical expertise Students from the Global South often do not have access to the education and mentorship required to develop emerging technologies,such as generative AI.This can contribute to a lack of global representation among generative AI researchers and engineers,with potential downstream ef

84、fects of unintended algorithmic biases and discrimination in generative AI products.Local access to high-quality education and generative AI expertise is key to creating a sustainable talent pipeline and widening the locations where generative AI research is done.Further,more researchers and enginee

85、rs from the Global South will lead to more diversity in generative AI ideas,enhanced innovation and increased opportunities for local experts to build and wield generative AI with local issues in mind.Governance Institutional capacity and policy developmentEconomically disadvantaged countries often

86、lack the financial,political and technical resources needed to develop effective AI governance policies,and regulators within these jurisdictions remain severely underfunded.According to a 2023 study of 193 countries,114 countries,almost exclusively from the Global South,lack any national AI strateg

87、y.71Disparity in AI governance capabilities can reinforce existing power imbalances and hinder global participation in the benefits of generative AI.The absence of governance policies for data and AI can lead to privacy violations,potential misuse of AI and a missed opportunity to harness AI for pos

88、itive socioeconomic development,among others.Further,underfunded regulatory institutions may be ill-equipped to address the ethical,legal and social implications of AI.Generative AI Governance:Shaping a Collective Global Future11ConclusionThe global governance landscape for AI is complex,fragmented

89、and rapidly evolving,with new and amplified challenges presented by the advent of generative AI.To effectively harness the global opportunities of generative AI and address its associated risks,there is a critical need for international cooperation and jurisdictional interoperability.Coordinated mul

90、tistakeholder efforts,including government,civil society,academia,industry and impacted communities,are essential.As humans drive the development of this technology and policy,responses must be developed to increase equity and inclusion in the development of AI,including with the countries of the Gl

91、obal South.It is up to stakeholders to take concrete action on access and inclusion.The World Economic Forum and its AI Governance Alliance are committed to driving this change,using its unique platform as a catalyst to convene diverse voices from around the world and urge them to act on vital issue

92、s,promote shared learnings and advance novel solutions.Generative AI Governance:Shaping a Collective Global Future12ContributorsAcknowledgementsWorld Economic Forum Benjamin LarsenLead,Artificial Intelligence and Machine LearningCathy LiHead of AI,Data and Metaverse;Deputy Head,Centre for the Fourth

93、 Industrial Revolution;Member of the Executive CommitteeKarla Yee AmezagaLead,Data Policy and AIAI Governance Alliance Project FellowsArnab ChakrabortySenior Managing Director,Global Responsible AI Lead,AccentureRafi LazersonGenAI Policy Manager,AccentureValerie MorignatGlobal Responsible AI Lead fo

94、r Life Sciences,AccentureManal SiddiquiResponsible AI Manager,AccentureAli ShahGlobal Principal Director for Responsible AI,AccentureKathryn WhiteGlobal Principal Director for Innovation Incubation,AccentureThis paper is a combined effort based on numerous interviews,discussions,workshops and resear

95、ch.The opinions expressed herein do not necessarily reflect the views of the individuals or organizations involved in the project or listed below.Sincere thanks are extended to those who contributed their insights via interviews and workshops,as well as those not captured below.Sincere appreciation

96、is extended to the following working group members,who spent numerous hours providing critical input and feedback to the drafts.Their diverse insights are fundamental to the success of this work.Lovisa AfzeliusChief Executive Officer,Apriori BioHassan Al-DarbestiAdviser to the Minister and Director,

97、International Cooperation Department,Ministry of Information and Communication Technology(ICT)of QatarUthman AliSenior Product Analyst,AI Ethics SME,BPErich David AndersenGeneral Counsel;Head,Corporate Affairs,TikTokJason AndersonGeneral Counsel,Vice-President and Corporate Secretary,DataStaxNorbert

98、o AndradeProfessor and Academic Director,IE UniversityRichard BenjaminsChief AI and Data Strategist,TelefonicaSaqr BinghalibExecutive Director,Artificial Intelligence,Digital Economy and Remote Work Applications Office,United Arab EmiratesAnu BradfordProfessor of Law,Columbia Law SchoolMichal Brand-

99、GoldVice-President General Counsel,ActivefenceGenerative AI Governance:Shaping a Collective Global Future13Adrian BrownExecutive Director,Center for Public ImpactWinter CaseySenior Director,SAPSimon ChestermanSenior Director of AI Governance,AI Singapore,National University of SingaporeMelinda Clayb

100、aughDirector,Privacy Policy,Meta PlatformsAmanda CraigSenior Director,Responsible AI Public Policy,MicrosoftRene CummingsData Science Professor and Data Activist in Residence,University of VirginiaNicholas DirksPresident and Chief Executive Officer,The New York Academy of SciencesNita FarahanyRobins

101、on O.Everett Professor of Law and Philosophy;Director,Duke Science and Society,Duke UniversityMax FenkellVice-President,Government Relations,Scale AIKay Firth-ButterfieldSenior Research Fellow,University of Texas at AustinKatharina FreyDeputy Head,Digitalisation Division,Federal Department of Foreig

102、n Affairs,Federal Department of Foreign Affairs(FDFA)of SwitzerlandAlice FriendHead,Artificial Intelligence and Emerging Tech Policy,GoogleTony GaffneyChief Executive Officer,Vector InstituteEugenio GarciaDeputy Consul-General,San Francisco,Ministry of Foreign Affairs of BrazilUrs GasserDean,TUM Sch

103、ool of Social Sciences and Technology,Technical University of MunichAvi GesserPartner,Debevoise&PlimptonDebjani GhoshPresident,National Association of Software and Services Companies(NASSCOM)Danielle Gilliam-MooreDirector,Global Public Policy,SalesforceBrian GreenDirector,Technology Ethics,Santa Cla

104、ra UniversitySamuel GregoryExecutive Director,WITNESSKoiti HasidaDirector,Artificial Intelligence in Society Research Group,RIKEN Center for Advanced Intelligence Project,RIKENDan HendrycksExecutive Director,Center for AI SafetyBenjamin HughesSenior Vice-President,Artificial Intelligence(AI)&Real Wo

105、rld Data(RWD),IQVIADan JermynChief Decision Scientist,Commonwealth Bank of AustraliaJeff Jianfeng CaoSenior Research,Tencent Research InstituteSam KaplanAssistant General Counsel,Public Policy&Government Affairs,Palo Alto NetworksKathryn KingGeneral Manager,Technology&Strategy,Office of the eSafety

106、Commissioner,AustraliaEdward S.KnightExecutive Vice-Chairman,NasdaqAndrew JP LevyChief Corporate and Government Affairs Officer,AccentureCaroline LouveauxChief Privacy and Data Responsibility Officer,MastercardShawn MaherGlobal Vice-Chair,Public Policy,EYGevorg MantashyanFirst Deputy Minister of Hig

107、h-Tech Industry,Ministry of High-Tech Industry of ArmeniaGary MarcusChief Executive Officer,Center for Advancement of Trustworthy AIGregg MelinsonSenior Vice-President,Corporate Affairs,Hewlett Packard EnterpriseNicolas MiailheFounder and President,The Future Society(TFS)Robert MiddlehurstSenior Vic

108、e-President,Regulatory Affairs,e&InternationalGenerative AI Governance:Shaping a Collective Global Future14Casey MockChief Policy and Public Affairs Officer,Center for Humane TechnologyChandler MorseVice-President,Corporate Affairs,WorkdayMiho NaganumaSenior Executive Professional,Digital Trust Busi

109、ness Strategy Department,NECDan NechitaHead,Cabinet,MEP Drago Tudorache,European ParliamentMichael NunesHead,Government Advisory,VisaBo Viktor NylundDirector,UNICEF Innocenti Global Office of Research and Foresight,United Nations Childrens Fund(UNICEF)Madan OberoiExecutive Director,Technology and In

110、novation,International Criminal Police Organization(INTERPOL)Michael OrtizSenior Director,Policy,Sequoia Capital OperationsFlorian OstmannHead,AI Governance and Regulatory Innovation,The Alan Turing InstituteMarc-Etienne OuimetteLead,Global AI Policy,Amazon Web ServicesTimothy PersonsPrincipal,Digit

111、al Assurance and Transparency of US Trust Solutions,PwCTiffany PhamFounder and Chief Executive Officer,MogulValerie PisanoPresident and Chief Executive Officer,MILA,Quebec Artificial Intelligence InstituteOreste PollicinoProfessor,Constitutional Law,Bocconi UniversityCatherine QuinlanVice-President,

112、AI Ethics,IBMMartin RauchbauerCo-Director and Founder,Tech Diplomacy NetworkAlexandra Reeve GivensChief Executive Officer,Center for Democracy and TechnologyPhilip ReinerChief Executive Officer,Institute for Security and TechnologyAndrea RendaSenior Research Fellow,Centre for European Policy Studies

113、(CEPS)Sam RizzoHead,Global Policy Development,Zoom Video CommunicationsJohn RoeseGlobal Chief Technology Officer,Dell TechnologiesArianna RufiniICT Adviser to the Minister,Ministry of Enterprises and Made in ItalyCrystal RugegeManaging Director,Centre for the Fourth Industrial Revolution,RwandaNayat

114、 Sanchez-PiChief Executive Officer,INRIA ChileThomas SchneiderAmbassador,Director of International Affairs,Swiss Federal Office of Communications,Federal Department of the Environment,Transport,Energy and Communications(DETEC)Robyn ScottCo-Founder and Chief Executive Officer,ApoliticalVar ShankarDir

115、ector,Policy,Responsible Artificial Intelligence InstituteNavrina SinghFounder and Chief Executive Officer,Credo AIIrina SoeffkyDirector,National,European and International Digital Policy,Federal Ministry for Digital and Transport of GermanyUyi StewartChief Data and Technology Officer,data.orgChizur

116、u SugaDirector,Digital Economy,Ministry of Economy,Trade and Industry of JapanArun SundararajanHarold Price Professor,Entrepreneurship and Technology,Stern School of Business,New York UniversityNabiha SyedChief Executive Officer,The MarkupPatricia ThaineCo-Founder and Chief Executive Officer,Private

117、 AIV Valluvan VelooDirector,Manufacturing Industry,Science and Technology Division,Ministry of Economy,MalaysiaGenerative AI Governance:Shaping a Collective Global Future15Rishi VarmaSenior Vice-President and General Counsel,Hewlett Packard EnterpriseOtt VelsbergGovernment Chief Data Officer,Ministr

118、y of Economic Affairs and Information Technology of EstoniaMiriam VogelPresident and Chief Executive Officer,Equal AIArif ZeynalovTransformation Chief Information Officer,Ministry of Economy of the Republic of AzerbaijanWorld Economic ForumJohn BradleyLead,Metaverse InitiativeKaryn GormanCommunicati

119、ons Lead,Metaverse InitiativeDevendra JainLead,Artificial Intelligence,Quantum TechnologiesJenny JoungSpecialist,Artificial Intelligence and Machine LearningDaegan KingeryEarly Careers Programme,AI Governance AllianceConnie KuangLead,Generative AI and Metaverse Value CreationHannah RosenfeldSpeciali

120、st,Artificial Intelligence and Machine LearningSupheakmungkol SarinHead,Data and Artificial Intelligence EcosystemsStephanie TeeuwenSpecialist,Data and AIHesham ZafarLead,Digital TrustAccenturePatrick ConnollyResearch ManagerCharlie MoskowitzSenior Manager,Government RelationsAnna SchillingData&AI S

121、trategy ManagerSekhar TewariAssociate Research ManagerDikshita VenkateshResearch Senior Analyst,Responsible AIJapan External Trade OrganizationGenta AndoExecutive Director;Project Fellow,World Economic ForumProductionLaurence Denmark Creative Director,Studio MikoSophie Ebbage Designer,Studio MikoMar

122、tha Howlett Editor,Studio MikoGenerative AI Governance:Shaping a Collective Global Future16Endnotes1.World Economic Forum,Unlocking value from Generative AI:Guidance for responsible transformation,2024.2.“Generative AI could raise global GDP by 7%”,Goldman Sachs,05 April 2023,https:/ Economic Forum,

123、Toolkit for Digital Safety Design Interventions and Innovations:Typology of Online Harms,2023,https:/www3.weforum.org/docs/WEF_Typology_of_Online_Harms_2023.pdf.4.Schaake,Marietje,“There can be no AI regulation without corporate transparency”,Financial Times,31 October 2023 https:/cyber.fsi.stanford

124、.edu/publication/there-can-be-no-ai-regulation-without-corporate-transparency.5.Appel,Gil,Juliana Neelbauer and David A.Schweidel,“Generative AI Has an Intellectual Property Problem”,Harvard Business Review,7 April 2023,https:/hbr.org/2023/04/generative-ai-has-an-intellectual-property-problem.6.Thes

125、e approaches can be complementary.For example,a jurisdiction may decide to govern predictable risks with a risk-based approach,while leaving unpredictable risks governed by an outcomes-based approach.7.Council of the EU and the European Council,Artificial intelligence act:Council and Parliament stri

126、ke a deal on the first rules for AI in the world Press release,9 December 2023,https:/www.consilium.europa.eu/en/press/press-releases/2023/12/09/artificial-intelligence-act-council-and-parliament-strike-a-deal-on-the-first-worldwide-rules-for-ai/.8.“The Artificial Intelligence and Data Act(AIDA)Comp

127、anion document”,Government of Canada,2023,https:/ised-isde.canada.ca/site/innovation-better-canada/en/artificial-intelligence-and-data-act-aida-companion-document.9.“Committee of jurists approves text with rules for artificial intelligence”,Senado Noticias,1 December 2022,https:/www12.senado.leg.br/

128、noticias/materias/2022/12/01/comissao-de-juristas-aprova-texto-com-regras-para-inteligencia-artificial.10.“Legal Alert:Chile takes first steps towards regulation of Artificial Intelligence”,DLA PIPER,15 June 2023,https:/www.dlapiper.cl/en/2023/06/15/legal-alert-chile-takes-first-steps-towards-regula

129、tion-of-artificial-intelligence/.11.Republic of the Phillipines,House Bill 7396,1 March 2023,https:/hrep-website.s3.ap-southeast- AI Regulation Dilemma”,The Diplomat,27 October 2023,https:/ Parliament,Artificial Intelligence Act:deal on comprehensive rules for trustworthy AI Press release,9 December

130、 2023,https:/www.europarl.europa.eu/news/en/press-room/20231206IPR15699/artificial-intelligence-actdeal-on-comprehensive-rules-for-trustworthy-ai.14.Government of Singapore,AI for the Public Good For Singapore and the World,2023,https:/file.go.gov.sg/nais2023.pdf.15.Malaysia Ministry of Science,Tech

131、nology&Innovation(MOSTI),Malaysia National Artificial Intelligence Roadmap 2021-2025,August 2022 https:/airmap.my/wp-content/uploads/2022/08/AIR-Map-Playbook-final-s.pdf16.National Strategy for Data&AI(NSDAI),Kingdom of Saudi Arabia,Realizing our Best Tomorrow,2020,https:/ai.sa/Brochure_NSDAI_Summit

132、%20version_EN.pdf.17.Cabinet Office,Government of Japan,AI Strategy 2022,2022,https:/www8.cao.go.jp/cstp/ai/aistrategy2022_gaiyo.pdf.18.Republic of Rwanda Ministry of ICT and Innovation,The National AI Policy,2022,https:/rura.rw/fileadmin/Documents/ICT/Laws/Rwanda_national_Artificial_intelligence_Po

133、licy.pdf.19.For a live repository of over 1,000 AI policy initiatives see:OECD.AI Policy Observatory,National AI policies&strategies Infographic and live repository,https:/oecd.ai/en/dashboards/overview/policy.20.World Economic Forum,Unlocking Value from Generative AI:Guidance for Responsible Transf

134、ormation,2024.21.“NIST Seeks Collaborators for Consortium Supporting Artificial Intelligence Safety”,National Institute of Standards and Technology(NIST),2 November 2023,https:/www.nist.gov/news-events/news/2023/11/nist-seeks-collaborators-consortium-supporting-artificial-intelligence.22.World Econo

135、mic Forum,Data Equity:Foundational Concepts for Generative AI,2023,pp.10,https:/www3.weforum.org/docs/WEF_Data_Equity_Concepts_Generative_AI_2023.pdf.23.World Economic Forum,Jobs of Tomorrow:Large Language Models and Jobs,2023,https:/www.weforum.org/publications/jobs-of-tomorrow-large-language-model

136、s-and-jobs/.24.Henderson,Peter,Xuechen Li,Dan Jurafsky,Tatsunori Hashimoto,et al.,“Foundation Models and Copyright Questions”,Stanford University Human-Centered Artificial Intelligence,November 2023,https:/hai.stanford.edu/policy-brief-foundation-models-and-copyright-questions;see also:DAuria,Giusep

137、pina and Arun Sundararajan,“Rethinking Intellectual Property Law in an Era of Genarative AI”,TechREG Chronicle,November 2023,pp.3-11,https:/ Position on Foundation Models and Generative AI for the EU AI Acts Trilogue Negotiations,2023.26.World Economic Forum,Presidio AI Framework:Towards Safe Genera

138、tive AI Models,2024.Generative AI Governance:Shaping a Collective Global Future1727.Leibowicz,Claire,”Why watermarking AI-generated content wont guarantee trust online”,MIT Technology Review,9 August 2023,https:/ in-depth analysis on data equity and generative AI see:World Economic Forum,Data Equity

139、:Foundational Concepts for Generative AI,2023,https:/www.weforum.org/publications/data-equity-foundational-concepts-for-generative-ai/.29.Talat,Zeerak,Aurlie Nvol,Stella Biderman,Miruna Clinciu,et al.,“You reap what you sow:On the Challenges of Bias Evaluation Under Multilingual Settings”,in Proceed

140、ings of BigScience Episode#5 Workshop on Challenges&Perspectives in Creating Large Language Models,eds.Angela Fan,Suzana Ilic,Thomas Wolf and Matthias Gall,Association for Computational Linguistics,2022,pp.26-41.30.Treat,David and Marie Wallace,“3 urgent questions to ask as we navigate a new digital

141、 identity”,World Economic Forum,28 September 2023,https:/www.weforum.org/agenda/2023/09/3-urgent-questions-digital-identity/.31.Hendrycks,Dan,Mantas Mazeika and Thomas Woodside,“An overview of catastrophic ai risks”,arXiv,9 October 2023,https:/arxiv.org/pdf/2306.12001.pdf.32.For a live crowd-sourced

142、 repository of AI-related harms see:Artificial Intelligence Incident Database(AIID)live repository,AI Incident Database,https:/incidentdatabase.ai/.33.Center for AI Safety,Statement on AI Risk,2023,https:/www.safe.ai/statement-on-ai-risk#open-letter.34.Bengio,Yoshua,Geoffrey Hinton,Andrew Yao,Dawn S

143、ong,et al.,“Managing AI Risks in an Era of Rapid Progress”,arXiv,2023,https:/arxiv.org/pdf/2310.17688.pdf.35.Frank,Michael,“Managing Existential Risk from AI without Undercutting Innovation”,Center for Strategic and International Studies(CSIS),10 July 2023,https:/www.csis.org/analysis/managing-exist

144、ential-risk-ai-without-undercutting-innovation.36.Thornhill,John,“AI will never threaten humans,says top Meta scientist”,Financial Times,18 October 2023,https:/ 12”,Unmasking AI,Penguin Random House,2023.38.Gebru,Timnit,Emily M.Bender,Angelina McMillan-Major and Margaret Mitchell,“Statement from the

145、 listed authors of Stochastic Parrots on the“AI pause”letter”,DAIR Institute,31 March 2023,https:/www.dair-institute.org/blog/letter-statement-March 2023/.39.Hanna,Alex and Emily M.Bender,“AI Causes Real Harm.Lets Focus on That over the End-of-Humanity Hype”,Scientific American,12 August 2023,https:

146、/ One is Immune to AI Harms with Dr.Joy Buolamwini”,Your Undivided Attention podcast transcript,episode 77,26 October 2023,https:/assets-global.website- Balance in Generative AI:Open-Source Versus Proprietary Models”,Tencent,19 October 2023,https:/ vs.Closed-Source AI”,Stanford eCorner,26 April 2023

147、,https:/ecorner.stanford.edu/clips/open-source-vs-closed-source-ai/.42.World Economic Forum,Presidio AI Framework:Towards Safe Generative AI Models,2024.43.Mozilla,Joint Statement on AI Safety and Openness,31 October 2023,https:/open.mozilla.org/letter/.44.Harris,David Evan,“How to Regulate Unsecure

148、d“Open-Source”AI:No Exemption”,Tech Policy Press,3 December 2023,https:/www.techpolicy.press/how-to-regulate-unsecured-opensource-ai-no-exemptions/.45.Sutskever,Ilya,“Open-Source vs.Closed-Source AI”,Stanford eCorner,26 April 2023,https:/ecorner.stanford.edu/clips/open-source-vs-closed-source-ai/.46

149、.Harris,David Evan,“How to Regulate Unsecured“Open-Source”AI:No Exemption”,Tech Policy Press,3 December 2023,https:/www.techpolicy.press/how-to-regulate-unsecured-opensource-ai-no-exemptions/.47.Sutskever,Ilya,“Open-Source vs.Closed-Source AI”,Stanford eCorner,26 April 2023,https:/ecorner.stanford.e

150、du/clips/open-source-vs-closed-source-ai/.48.“OECD Artifical Intelligence Papers”,OECD Library,n.d.,https:/www.oecd-ilibrary.org/science-and-technology/common-guideposts-to-promote-interoperability-in-ai-risk-management_ba602d18-en.49.“ASEAN initiates regional discussion on generative AI Policy”,Ass

151、ociation of Southeast Asian Nations,7 December 2023,https:/asean.org/asean-initiates-regional-discussion-on-generative-ai-policy/.50.Ministry of Communications and Information,The 3rd ASEAN Digital Ministers Meeting and Related Meetings at the Philippines Press release,9 February 2023,https:/www.mci

152、.gov.sg/media-centre/press-releases/the-3rd-asean-digital-ministers-meeting-at-the-philippines/.51.“G7 Leaders Statement on the Hiroshima AI Process”,Eurpoean Commission,30 October 2023,https:/digital-strategy.ec.europa.eu/en/library/g7-leaders-statement-hiroshima-ai-process.52.For more examples of

153、international collaboration,see:Oxford Insights,2023 Government AI Readiness Index,2023,pp.9-10,https:/ AI Governance:Shaping a Collective Global Future1853.Organisation for Economic Co-operation and Development(OECD),G7 Hiroshima Process on Generative Artificial Intelligence(AI),2023,https:/read.oe

154、cd-ilibrary.org/science-and-technology/g7-hiroshima-process-on-generative-artificial-intelligence-ai_bf3c0c60-en#page1.54.“AI Risk Management Framework”,National Institute of Standards and Technology(NIST),n.d.,https:/www.nist.gov/itl/ai-risk-management-framework.55.“Standardisation Strategy Consult

155、ation-Feedback From ETUC”,European Trade Union Confederation(ETUC),28 July 2021,https:/ec.europa.eu/info/law/better-regulation/have-your-say/initiatives/13099-Standardisation-strategy/F2663296_en.56.“ISO/IEC JTC 1/SC 42”,International Organization for Standardization(ISO),2017,https:/www.iso.org/com

156、mittee/6794475.html.57.EU:The EU AI Act will also rely on compliance with harmonized standards aligned with international standardization efforts as a means to demonstrate conformity with its requirements.US:The long-standing Circular No.A-119 on federal development and use of voluntary consensus st

157、andards and conformity assessment outlines a commitment to using international standards whenever possible.China:2021 National Standardization Development Outline reiterates Beijings investment in AI standards and conformity assessment,laying out standards for AI development and deployment,and align

158、ing these standards with international ones.58.Infocomm Media Development Authority(IMDA),First of its kind Generative AI Evaluation Sandbox for Trusted AI by AI Verify Foundation and IMDA Press release,31 October 2023,https:/www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-rel

159、eases/2023/generative-ai-evaluation-sandbox.59.United Arab Emirates Government,“Regulatory Sandboxes in the UAE”,n.d.,https:/u.ae/en/about-the-uae/digital-uae/regulatory-framework/regulatory-sandboxes-in-the-uae.60.“ANPDs Call for Contributions to the regulatory sandbox for artificial intelligence a

160、nd data protection in Brazil is now open”,Autoridade Nacional de Proteo de Dados,3 October 2023,https:/www.gov.br/anpd/pt-br/assuntos/noticias/anpds-call-for-contributions-to-the-regulatory-sandbox-for-artificial-intelligence-and-data-protection-in-brazil-is-now-open.61.“Regulatory Sandbox”,Informat

161、ion Commissioners Office,n.d.,https:/ico.org.uk/sandbox.62.European Parliament,Artificial Intelligence Act:deal on comprehensive rules for trustworthy AI Press release,9 December 2023,https:/www.europarl.europa.eu/news/en/press-room/20231206IPR15699/artificial-intelligence-act-deal-on-comprehensive-

162、rules-for-trustworthy-ai.63.Ministry of Public Service,Administrative and Institutional Reforms,Sandbox Framework for Adoption of Innovative Technologies in the Public Service,2021,https:/civilservice.govmu.org/Documents/Circulars%202021/Booklet%20Sandbox%20framework.pdf.64.Brookings Institution,“Wh

163、y the Global South has a stake in dialogues on AI governance”,YouTube,23 October 2023.https:/ T.,“AI in the Global South:Opportunities and challenges towards more inclusive governance”,Brookings,1 November 2023,https:/www.brookings.edu/articles/ai-in-the-global-south-opportunities-and-challenges-tow

164、ards-more-inclusive-governance/.66.African Union High-Level Panel on Emerging Technologies,AI for Africa:Artificial Intelligence for Africas Socio-Economic Development,2021,https:/www.nepad.org/publication/ai-africa-artificial-intelligence-africas-socio-economic-development.67.Li,Pengfei,Jianyi Yang

165、,Mohammad A.Islam and Shaolei Ren,“Making AI Less“Thirsty”:Uncovering and Addressing the Secret Water Footprint of AI Models”,arXiv,29 October 2023,https:/arxiv.org/pdf/2304.03271.pdf.68.OECD,AI language models:Technological,socio-economic and policy considerations,2023,https:/doi.org/10.1787/13d38f

166、92-en.69.Ludvigsen,Kasper Groes Albin,“The Carbon Footprint of ChatGPT”,Towards Data Science,21 December 2022,https:/ AI Through Acquisition and Procurement,14 September 2023,Testimony presented to the U.S.Senate Committee on Homeland Security and Governmental Affairs,Washington DC.https:/hai.stanfo

167、rd.edu/sites/default/files/2023-09/Fei-Fei-Li-Senate-Testimony.pdf.71.Oxford Insights,2023 Government AI Readiness Index,2023,https:/ by EC/OECD(2021),database of national AI policies,https:/oecd.ai.73.Department of National Planning,Republic of Colombia,Poltica Nacional paralaTransformacin Digital

168、e Inteligencia Artificial,2019,https:/colaboracion.dnp.gov.co/CDT/Conpes/Econmicos/3975.pdf.74.Shimoda Uechi,Cristina Akemiand Thiago Guimares Moraes,“Brazils path to responsible AI”,OECD,27 July 2023,https:/oecd.ai/en/wonk/brazils-path-to-responsible-ai.75.Mauritius Working Group on AI,Mauritius Ar

169、tificial Intelligence Strategy,2018,https:/ncb.govmu.org/ncb/strategicplans/MauritiusAIStrategy2018.pdf.76.Republic of Rwanda Ministry of ICT and Innovation,The National AI Policy,2022,https:/www.minict.gov.rw/index.php?eID=dumpFile&t=f&f=67550&token=6195a53203e197efa47592f40ff4aaf24579640e.Generati

170、ve AI Governance:Shaping a Collective Global Future1977.Sierra Leone Directorate of Science Technology&Innovation,Sierra Leone National Innovation&Digital Strategy,2019,https:/www.dsti.gov.sl/wp-content/uploads/2019/11/Sierra-Leone-National-Innovation-and-Digital-Strategy.pdf.78.Government of the So

171、cialist Republic of Viet Nam,National Strategy on R&D and Application of Artificial Intelligence,2021,https:/wp.oecd.ai/app/uploads/2021/12/Vietnam_National_Strategy_on_RD_and_Application_of_AI_2021-2030.pdf.79.“AI Towards Indonesias Vision 2045”,Indonesia Center for Artificial Intelligence Innovati

172、on,n.d.,https:/ai-innovation.id/strategi.80.Thomson Reuters Foundation,AI Governance for Africa,Part 1 and 2,2023,https:/www.trust.org/dA/97390870db/pdfReport/AI%20Governance%20for%20Africa%20Toolkit%20-%20Part%201%20and%202.pdf.81.Chatham House,Reflections on Building More Inclusive Global Governan

173、ce,2021,https:/www.chathamhouse.org/sites/default/files/2021-04/2021-04-15-reflections-building-inclusive-global-governance.pdf.Generative AI Governance:Shaping a Collective Global Future20World Economic Forum9193 route de la CapiteCH-1223 Cologny/GenevaSwitzerland Tel.:+41(0)22 869 1212Fax:+41(0)22

174、 786 2744contactweforum.orgwww.weforum.orgThe World Economic Forum,committed to improving the state of the world,is the International Organization for Public-Private Cooperation.The Forum engages the foremost political,business and other leaders of society to shape global,regional and industry agendas.