This is the sec\u00adond post in our blog series \u201cAI in the uni\u00adver\u00adsity<\/strong>\u201d. After the first post was about AI detec\u00adtors<\/a><\/em> in the first post, in this post we looked at AI and OER from a legal per\u00adspec\u00adtive<\/strong><\/em>. This will soon be fol\u00adlowed by an arti\u00adcle on prompt tips for high-qual\u00adity OER<\/em>. We will then con\u00adclude the series with manda\u00adtory AI skills<\/em> in higher edu\u00adca\u00adtion.<\/p>\n\n\n\n Recently, in the atten\u00adtion econ\u00adomy, the topic of \u201cAI\u201d has almost sup\u00adplanted the topic of \u201cOER\u201d in the uni\u00adver\u00adsity con\u00adtext. Every\u00adone is talk\u00ading about AI: AI projects are being funded, AI work\u00adshops are fully booked. OER has moved down the list of pri\u00ador\u00adi\u00adties. These devel\u00adop\u00adments are under\u00adstand\u00adable. And yet it should not be for\u00adgot\u00adten that not only AI, but also OER<\/em> has great poten\u00adtial for teach\u00ading<\/em>. So why not com\u00adbine the two top\u00adics to make the most of their poten\u00adtial<\/a>? <\/p>\n\n\n\n The inter\u00adac\u00adtion between AI and OER can bring con\u00adsid\u00ader\u00adable added value for teach\u00ading: AI can pre\u00adpare open edu\u00adca\u00adtional mate\u00adri\u00adals in a didac\u00adti\u00adcally clever and appeal\u00ading way, struc\u00adture them and much more. In turn, OER can help to ensure that AI pro\u00adgrams are trained with high-qual\u00adity mate\u00adri\u00adals and thus deliver more reli\u00adable results. <\/p>\n\n\n\n And what does the law say about this rela\u00adtion\u00adship? To what extent is a legally com\u00adpli\u00adant use of AI pro\u00adgrams in OER even pos\u00adsi\u00adble? We will try to answer this ques\u00adtion here. We will only address OER-rel\u00ade\u00advant aspects (exclud\u00ading data pro\u00adtec\u00adtion aspects). Please also note that legal devel\u00adop\u00adments in the field of AI are very dynamic. The expla\u00adna\u00adtions in this blog post refer to the cur\u00adrent legal sit\u00adu\u00ada\u00adtion (Octo\u00adber 2024). <\/p>\n\n\n\n AI out\u00adput (e.g. images, texts, info\u00adgraph\u00adics, etc.) is in the pub\u00adlic domain under Ger\u00adman copy\u00adright law<\/strong>, i.e. free from copy\u00adright. This is because, accord\u00ading to the clear word\u00ading of Sec\u00adtion 2 (2) UrhG, only humans can be the cre\u00adators of an intel\u00adlec\u00adtual out\u00adput. Con\u00adse\u00adquently, the AI can\u00adnot have any copy\u00adrights to the out\u00adput. <\/p>\n\n\n\n Oper\u00ada\u00adtors of AI pro\u00adgrams as legal enti\u00adties are not con\u00adsid\u00adered authors either. The author\u00adship of the AI out\u00adput does not lie with the pro\u00adgram\u00admers of the AI pro\u00adgram either, because AI is gen\u00ader\u00adally not con\u00adtrol\u00adlable. It func\u00adtions autonomously. The same applies to users of the pro\u00adgram. It is the\u00ado\u00adret\u00adi\u00adcally con\u00adceiv\u00adable to manip\u00adu\u00adlate AI out\u00adput with very clever prompts (com\u00admands). How\u00adever, it will gen\u00ader\u00adally not be pos\u00adsi\u00adble to con\u00adtrol the AI entirely with these. As a result, although the smart prompt is pro\u00adtected by copy\u00adright as the user\u2019s work, the out\u00adput remains in the pub\u00adlic domain. <\/p>\n\n\n\n Some\u00adthing else may apply if the AI is only used as a tool<\/strong> in the cre\u00adative process (i.e. only as a sup\u00adport), but the actual per\u00adfor\u00admance can be objec\u00adtively attrib\u00aduted to a human being. This can be com\u00adpared to the use of an image edit\u00ading pro\u00adgram. In this case, the copy\u00adright lies with the cre\u00adator of the work. <\/p>\n\n\n\n Copy\u00adright pro\u00adtec\u00adtion can also arise if the AI prod\u00aduct is exten\u00adsively cre\u00adatively reworked by a human<\/strong>. The result of the adap\u00adta\u00adtion must be suf\u00adfi\u00adciently indi\u00advid\u00adual, i.e. take on the per\u00adsonal char\u00adac\u00adter\u00adis\u00adtics of the cre\u00adator of the work. This is not the case if only col\u00adors or sizes are adjusted or addi\u00adtions are made. Whether the required indi\u00advid\u00adu\u00adal\u00adity and cre\u00adativ\u00adity (level of cre\u00adativ\u00adity) are present always depends on the spe\u00adcific indi\u00advid\u00adual case. You can find out more about this in the arti\u00adcle AI and OER: How well do they go together?<\/a> by Georg Fis\u00adcher. <\/p>\n\n\n\n As the pure AI out\u00adput is in the pub\u00adlic domain, it can gen\u00ader\u00adally be used in OER mate\u00adri\u00adals. How\u00adever, this requires that no third-party rights (e.g. copy\u00adright, per\u00adsonal rights, trade\u00admark rights, etc.) are infringed by the out\u00adput. <\/p>\n\n\n\n The pub\u00adli\u00adca\u00adtion of infring\u00ading con\u00adtent under an open license can be warned. This is because copy\u00adright law does not pro\u00adtect the good faith belief that the con\u00adtent used does not infringe the prop\u00aderty rights of third par\u00adties. Rather, users have a duty to check the con\u00adtent used for infringe\u00adments<\/strong>, which is not easy with AI-gen\u00ader\u00adated con\u00adtent. Users can\u00adnot know which sources are used to train the AI and whether the rights of third par\u00adties have been infringed. The\u00ado\u00adret\u00adi\u00adcally, it is even pos\u00adsi\u00adble for the AI to repro\u00adduce third-party works with which it has been trained in its out\u00adput one-to-one or in part, but this is more of an excep\u00adtion. <\/p>\n\n\n\n Most AI pro\u00adgrams do not copy other peo\u00adple\u2019s works, but cre\u00adate pat\u00adterns based on train\u00ading data, sim\u00adi\u00adlar to the human brain. The prob\u00ada\u00adbil\u00adity of being warned is there\u00adfore not par\u00adtic\u00adu\u00adlarly high. Nev\u00ader\u00adthe\u00adless, it is impor\u00adtant not to accept the gen\u00ader\u00adated result with\u00adout reflec\u00adtion, but to check for legal infringe\u00adments, e.g. with pla\u00adgia\u00adrism soft\u00adware <\/strong> The Euro\u00adpean AI Reg\u00adu\u00adla\u00adtion(AI Act<\/a>), which came into force on August 1, 2024, requires oper\u00ada\u00adtors of gen\u00ader\u00ada\u00adtive AI pro\u00adgrams to develop a copy\u00adright com\u00adpli\u00adance strat\u00adegy and pub\u00adlish detailed sum\u00admaries of train\u00ading data (recitals 106 and 107). How\u00adever, it will take some time for AI oper\u00ada\u00adtors to ful\u00adfill this oblig\u00ada\u00adtion, as the reg\u00adu\u00adla\u00adtion is being imple\u00admented grad\u00adu\u00adally. <\/p>\n\n\n\n Cau\u00adtion is also required when for\u00admu\u00adlat\u00ading prompts. No prompts<\/strong> should be pub\u00adlished that con\u00adtain copy\u00adrighted works or parts thereof, third-party brands, per\u00adsonal pho\u00adtos or other per\u00adsonal data<\/strong>. This can also result in a warn\u00ading. <\/p>\n\n\n\n In order for the AI to be able to gen\u00ader\u00adate con\u00adtent itself, it must first be trained with open sources (i.e. also with OER). This is prob\u00ada\u00adbly legally per\u00admis\u00adsi\u00adble in Ger\u00admany. It is pre\u00addom\u00adi\u00adnantly argued that the copy\u00adright bar\u00adrier for text and data min\u00ading in Sec\u00adtion 44 b UrhG jus\u00adti\u00adfies the train\u00ading. <\/p>\n\n\n\n Train\u00ading AI with third-party con\u00adtent is not so easy to pre\u00advent. Many read\u00aders will remem\u00adber the case involv\u00ading Meta<\/a> in the sum\u00admer of 2024. The NRW con\u00adsumer advice cen\u00adter had issued a warn\u00ading to Meta because the com\u00adpany wanted to use user data on Insta\u00adgram and Face\u00adbook to train AI with\u00adout the users\u2019 con\u00adsent. Ulti\u00admately, the users were allowed to object to the train\u00ading. How\u00adever, the way to object was some\u00adwhat com\u00adplex. <\/p>\n\n\n\n For con\u00adtent on web\u00adsites and blogs, it is pos\u00adsi\u00adble to declare a reser\u00adva\u00adtion of use against the train\u00ading<\/strong>, the so-called opt-out from data min\u00ading. This must be done by stor\u00ading a file in the root direc\u00adtory of the domain, i.e. in machine-read\u00adable form, so that it can be read by web crawlers. Web crawlers are pro\u00adgrams that search the Inter\u00adnet and ana\u00adlyze web\u00adsites. How exactly this works is explained, for exam\u00adple, in the tuto\u00adr\u00adial How to block Ope\u00adnAI\u2019s Chat\u00adGPT, Google\u2019s Gem\u00adini and other bots that want to use your texts for their AI<\/a> by Kai Spriesters\u00adbach. How\u00adever, an opt-out can also result in web\u00adsite con\u00adtent not being dis\u00adplayed as search results. <\/p>\n\n\n\n Since AI-gen\u00ader\u00adated con\u00adtent is in the pub\u00adlic domain, there is cur\u00adrently no legal oblig\u00ada\u00adtion to cite the source. Nev\u00ader\u00adthe\u00adless, mate\u00adri\u00adals cre\u00adated with AI should always be labeled as such. This makes it eas\u00adier for sub\u00adse\u00adquent users to under\u00adstand the cre\u00adation process and the ori\u00adgin of the mate\u00adr\u00adial. It is also made clear that the mate\u00adr\u00adial is in the pub\u00adlic domain, i.e. not sub\u00adject to any copy\u00adright restric\u00adtions. The label\u00ading could, for exam\u00adple, be as fol\u00adlows: \u201cAI-gen\u00ader\u00adated. Pub\u00adlic domain. Cre\u00adated with the pro\u00adgram XY. Prompt: XY. Edits: XY\u201d.<\/p>\n\n\n\n The OER mate\u00adr\u00adial in which AI-gen\u00ader\u00adated con\u00adtent is used can be openly licensed. Licenses should be cho\u00adsen that are the least restric\u00adtive for sub\u00adse\u00adquent use, i.e. CC Zero release, CC BY and CC BY SA. The pub\u00adlic domain AI con\u00adtent must be excluded from the license: \u201cThis OER mate\u00adr\u00adial is licensed under license XY. The marked AI-gen\u00ader\u00adated images are not cov\u00adered by the license. These are in the pub\u00adlic domain.\u201d <\/p>\n\n\n\n Some\u00adthing dif\u00adfer\u00adent applies to the so-called remix\u00ading (merg\u00ading) of mate\u00adri\u00adals into a new work. If AI-gen\u00ader\u00adated con\u00adtent and the OER mate\u00adr\u00adial are merged in such a way that a new work is cre\u00adated (e.g. an image col\u00adlage, a video), the new over\u00adall work may be openly licensed. Till Kreutzer explains this as an exam\u00adple in the video Open Edu\u00adca\u00adtional Resources, Copy\u00adright and AI<\/a>. <\/p>\n\n\n\n AI pro\u00adgrams can be used to cre\u00adate OER in com\u00adpli\u00adance with the law if the above rules are observed. If you would like to find out more about the topic or would like to deepen what you have read, visit our work\u00adshop AI & OER in use: Enhanc\u00ading OER in a diverse and legally com\u00adpli\u00adant way with AI<\/a>. For reg\u00adis\u00adtra\u00adtion and fur\u00adther ques\u00adtions, please con\u00adtact info@twillo.de.<\/a><\/p>\n\n\n\n <\/p>\n","protected":false},"excerpt":{"rendered":" This is the sec\u00adond post in our blog series \u201cAI in the uni\u00adver\u00adsity\u201d. After the first post was about AI detec\u00adtors in the first post, in this post we looked at AI and OER from a legal per\u00adspec\u00adtive. This will soon be fol\u00adlowed by an arti\u00adcle on prompt tips for high-qual\u00adity OER. We will then [\u2026]<\/p>\n","protected":false},"author":1,"featured_media":2968,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"wp_typography_post_enhancements_disabled":false,"footnotes":""},"categories":[145],"tags":[],"class_list":["post-2970","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-legal-issues"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/posts\/2970","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/comments?post=2970"}],"version-history":[{"count":0,"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/posts\/2970\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/media\/2968"}],"wp:attachment":[{"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/media?parent=2970"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/categories?post=2970"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.twillo.de\/en\/wp-json\/wp\/v2\/tags?post=2970"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}Is AI output protected by copyright?<\/h2>\n\n\n\n
Can AI-generated output be used in OER? What applies to prompts? <\/h2>\n\n\n\n
or a reverse search in the trade\u00admark reg\u00adis\u00adter. If pos\u00adsi\u00adble, the AI-gen\u00ader\u00adated result should be revised sev\u00aderal times with addi\u00adtional prompts or other pro\u00adgrams. <\/p>\n\n\n\nCan the AI be trained with copyrighted works? Can I prevent the AI from being trained with my works? <\/h2>\n\n\n\n
Do I have to label AI-generated content as such?<\/h2>\n\n\n\n
Conclusion<\/h2>\n\n\n\n