OECDpublishingINTELLECTUALPROPERTYISSUESINARTIFICIALINTELLIGENCETRAINEDONSCRAPEDDATAOECDARTIFICIALINTELLIGENCEPAPERSFebruary 2025No.33大GPAIOECDBETTERPOLICIESFORBETTERLIVES
2|NTELLECTUALPROPERTYISSUESINAITRAINEDONSCRAPEDDATAForewords.It provides anoverview ofthe role oftascrapingin-Altraining, curentlegal famewdconsideratins and potentalpolicyaproachesto help guide policymakers in navigating theseisues andacilitate a greater understanding of data scraping.This reportdisussedbyteOECDWorkingPartyonAlGovemance (AIGO)atitsNovember2023 andune 2024meetings. The Global Partnership on Atificial Intelligence (GPAl) discussed this work duringitsNovember2024PlenaryThe report was witen by Professor Lee Tiedich (Distinguished Faculty Fellow in Law and ResponsibleTechnology atDuke University), Karine Perset, andSaraFialho Esposito underthe supervision ofAudreyPlonk, Deputy Director of the OECD Science, Technology and Innovation DirectorateThis paper was approved and declassifed by witen procedure by the Global Partnership on AtifcalInteligence (GPAI) on 30 January 2025 and prepared for publication bythe OECD Secretariat.NotetoDelegations:ThisdocumentisalsoavailableonO.N.EMembers&PartnersunderthereferencecodeThis document, as well as any data and map included herein, are without pre judice to the status of onsovereignty over any territory, to the delimitation of international frontiers and boundaries and to the nameof any territory, dity or area.Coverimage:OKpargetr/Shutrstock.com@OECD2025CCAttbution4.0International(CCBY4.0)This workis m ade avalable under the Creative Comons Atribution 4.O nternational Icence.Bly using tis work, you aceptt6 be bound by te em s f hi ience htps Icreaivecormons oplicensesb/lA.0)Translationsiginal andaddthe following text:In the event of arvdiscrepancy betvweentheariginalwoik and the transato,only the ext oforgnal workshould be onsideredvaid
INTELLECTUALPROPERTYISSUES INAITRAINED ONSCRAPEDDATA|3ns-youmustdale the origial wok and ad the folowing text.This is an adaplation of anoriginalworkby theand arguments em ploy ed in his a aplaton shoul notb e repored as repesening the dficalviews oftheOECDorofitsMembercountiesThird-party material-the licencedoes not apply to third-party material in the work. If using such material, you areresponsible for obtaining permissionfrom the third party and for any claims of infingementYoumustnotusetheOECDlogo, visualidentity or coverimage without express permission or suagest the OECD endorsesunder this licence shall be settled by arbitration in accordance with the Permanent Court of AtbitrationI be Paris (France). The number of abitratorsshall be one
4/INTELLECTUALPROPERTYISSUESINAITRAINEDONSCRAPEDDATAAcknowledaementsGPAIworkshops on data scraping and intellectual propertytual PropertCommiteemeeting and the workshops.to thank variousstakeholder groups at the OECD for their valuable inputrly Nicole Primmer, Barry OErien, and Maylis Berille (Business at OECD-BIAC); Pam DixonInformation Socety Advisorv-CSISAC); and...