| ANNEX |
附属書 |
| to the Communication to the Commission Approval of the content of the draft Communication from the Commission –Explanatory Notice and Template for the Public Summary of Training Content for general-purpose AI models required by Article 53 (1)(d) of Regulation (EU) 2024/1689 (AI Act) |
委員会へのコミュニケーション委員会からのコミュニケーションの草案の内容の承認 – 規則 (EU) 2024/1689 (AI 法) 第 53 条 (1) (d) で要求される汎用 AI モデルの訓練内容に関する公開要約の説明文書およびテンプレート |
| 1. Background |
1. 背景 |
| (1) Regulation (EU) 2024/1689 of the European Parliament and the Council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending certain regulations[1] (‘the AI Act’) entered into force on 1 August 2024. Section 2 of Chapter V AI Act lays down harmonised rules for providers of general-purpose AI models, including obligations applicable to all providers of general-purpose AI models[2] and additional risk assessment and mitigation requirements for those of the most advanced general-purpose AI models posing systemic risks[3]. Those rules will apply as of 2 August 2025. |
(1) 2024年6月13日に欧州議会および理事会が採択した、人工知能に関する調和規則を定めるとともに、特定の規則を改正する規則(EU)2024/1689(以下「AI法」という)は、2024年8月1日に施行された。AI 法の第 V 章第 2 節では、汎用 AI モデル提供者に対する調和化された規則が定められており、これには、すべての汎用 AI モデル提供者に適用される義務[2]、および、システミックリスクをもたらす最先端の汎用 AI モデル提供者に対する追加的なリスクアセスメントおよび緩和要件[3] が含まれている。これらの規則は、2025 年 8 月 2 日から適用される。 |
| (2) Article 53(1)(d) AI Act requires all providers of general-purpose AI models to draw up and make publicly available a sufficiently detailed public summary of the content used for the training of the model (the ‘Summary’), according to a template provided by the AI Office (the ‘Template’). Recital 107 AI Act contains additional clarifications on the objectives of the Summary and the Template which include transparency on the data that is used for the training of general-purpose AI models, including text and data protected by copyright law. |
(2) AI 法第 53 条(1)(d) は、汎用 AI モデルのすべての提供者に対し、AI 事務局が提供するテンプレート(以下「テンプレート」という)に従って、モデルの訓練に使用した内容について、十分に詳細な公開要約(以下「要約」という)を作成し、公開することを義務付けています。前文 107 AI 法には、概要およびテンプレートの目的について、著作権法で保護されているテキストやデータを含む、汎用 AI モデルの訓練に使用されるデータの透明性など、追加の説明が記載されている。 |
| (3) Providers of all general-purpose AI models placed on the Union market must fulfil the above obligation, including providers of general-purpose AI models released under free and open-source licenses[4], in so far as the models fall within the scope of the AI Act[5]. Recital 107 AI Act contains additional clarifications on the Summary and the Template. |
(3) 連合市場に上市されるすべての汎用 AI モデル提供者は、そのモデルが AI 法[4] の適用範囲に該当する限り、無料かつオープンソースのライセンス[5] でリリースされる汎用 AI モデル提供者も含め、上記の義務を履行しなければならない。前文 107 AI 法には、概要およびテンプレートに関する追加の説明が含まれている。 |
| RELEVANT LEGAL TEXT |
関連法規 |
| Article 53(1)(d) AI Act. |
AI 法第 53 条(1)(d) |
| Providers of general-purpose AI models shall […] draw up and make publicly available a sufficiently detailed summary about the content used for training of the general-purpose AI model, according to a template provided by the AI Office. |
汎用 AI モデル提供者は、AI 事務局が提供するテンプレートに従って、汎用 AI モデルの訓練に使用される内容について、十分に詳細な概要を作成し、公開しなければならない。 |
| Recital 107 AI Act. |
前文 107 AI 法 |
| In order to increase transparency on the data that is used in the pre-training and training of general-purpose AI models, including text and data protected by copyright law, it is adequate that providers of such models draw up and make publicly available a sufficiently detailed summary of the content used for training the general-purpose AI model. While taking into due account the need to protect trade secrets and confidential business information, this summary should be generally comprehensive in its scope instead of technically detailed to facilitate parties with legitimate interests, including copyright holders, to exercise and enforce their rights under Union law, for example by listing the main data collections or sets that went into training the model, such as large private or public databases or data archives, and by providing a narrative explanation about other data sources used. It is appropriate for the AI Office to provide a template for the summary, which should be simple, effective, and allow the provider to provide the required summary in narrative form. |
著作権法で保護されているテキストやデータなど、汎用 AI モデルの事前学習および学習に使用されるデータの透明性を高めるため、当該モデルの提供者は、汎用 AI モデルの学習に使用される内容について、十分に詳細な要約を作成し、公開することが適切である。この概要は、営業秘密や機密の事業情報を保護する必要性を十分に考慮しつつ、著作権者を含む正当な利益を有する当事者が、例えば、大規模な民間または公的データベースやデータアーカイブなど、モデルの訓練に使用された主なデータコレクションまたはデータセットを記載し、その他のデータソースについて説明文を記載するなどして、EU 法に基づく権利を行使および執行することを容易にするため、技術的な詳細ではなく、一般的に包括的な内容とするべきだ。AI 事務局は、要約のテンプレートを提供することが適切であり、そのテンプレートは、簡潔かつ効果的であり、提供者が必要な要約を記述形式で提供できるものでなければならない。 |
| (4) Given that providers are obliged to draw up a Summary according to a Template provided by the Commission, the latter holds important legal value for the proper implementation of the AI Act. This Explanatory Notice and the Template annexed to it aim to help providers of general-purpose AI models comply with their obligation under Article 53(1)(d) of the AI Act in a simple, consistent and effective manner. |
(4) 提供者は、欧州委員会が提供するテンプレートに従って要約を作成する義務を負っているため、このテンプレートは AI 法を適切に実施する上で重要な法的価値を有する。この説明文書およびこれに付属するテンプレートは、汎用 AI モデル提供者が AI 法第 53 条(1)(d) の義務を、簡潔、一貫性、かつ効果的な方法で遵守することを支援することを目的としている。 |
| (5) The Template is based on the outcome of a multi-stakeholder consultation on general-purpose AI models, organised by the AI Office from 30 July to 18 September 2024[6]. Over 430 responses were submitted from a wide range of stakeholders. Based on this input, the AI Office prepared and presented its preliminary approach to the Template and allowed participants involved in the drawing up of the Code of Practice on General-Purpose AI[7] to provide additional written feedback. The current version of the Template annexed to this Explanatory Notice takes into account comments received from 111 stakeholders, including providers of general-purpose AI models, business associations, rightsholders organisations, academia, civil society and public authorities. The draft Template was also presented and discussed with the AI Board Steering subgroup on General-Purpose AI and with the European Parliament (IMCO-LIBE Committees) working group on AI. |
(5) テンプレートは、AI事務局が2024年7月30日から9月18日まで開催した汎用AIモデルに関する多者間協議の結果を基に作成された。幅広い関係者から 430 件以上の意見が寄せられた。この意見を踏まえ、AI 事務局はテンプレートの暫定的なアプローチを策定し、汎用 AI 行動規範[7] の策定に関与した関係者に対して、追加の書面によるフィードバックを求めることを決定した。この説明文書に添付されているテンプレートの現在のバージョンは、汎用 AI モデル提供者、事業者団体、権利者団体、学界、市民社会、公的機関など、111 のステークホルダーから寄せられたコメントを反映したものとなっている。また、テンプレートのドラフトは、AI 理事会汎用 AI に関する運営小委員会および欧州議会(IMCO-LIBE 委員会)の AI に関する作業部会にも提出され、議論された。 |
| 2. Objective of the Summary |
2. 要約の目的 |
| (6) General-purpose AI models are trained with large quantities of data for which there is typically limited information available. Recital 107 AI Act explains that the objective of the Summary is to increase transparency on the content used for the training of general-purpose AI models, including text and data protected by law and to facilitate parties with legitimate interests, including rightsholders, to exercise and enforce their rights under Union law. |
(6) 汎用 AI モデルは、通常、入手可能な情報が限られている大量のデータを用いて訓練される。前文 107 AI 法は、要約の目的は、法律で保護されているテキストやデータを含む、汎用 AI モデルの訓練に使用されるコンテンツの透明性を高め、権利者を含む正当な利益を有する当事者が、EU 法に基づく権利を行使および執行することを容易にするためであると説明している。 |
| (7) Such legitimate interests relate to copyright and related rights and other intellectual property rights, but also to other rights protected by Union law that should benefit from increased transparency. |
(7) このような正当な利益は、著作権および関連権利、その他の知的財産権だけでなく、透明性の向上によって恩恵を受けるべき、EU 法によって保護されるその他の権利にも関連している。 |
| (8) First, in relation to intellectual property rights, including copyright and related rights, transparency of the data used for the model training should help rightsholders obtain relevant information on the content used in the training of general-purpose AI models. This information is needed to facilitate the exercise of their fundamental right to intellectual property[8] and the fundamental right to an effective remedy in the enforcement of their rights, as provided for in Union law in the area of intellectual property rights. In the case of copyright and related rights, transparency of the training data will contribute to ensuring that general-purpose AI models providers comply with Union law on copyright and related rights[9]. |
(8) まず、著作権および関連権利を含む知的財産権に関しては、モデル訓練に使用されるデータの透明性は、権利者が汎用 AI モデルの訓練に使用されるコンテンツに関する関連情報を入手するのに役立つはずだ。この情報は、知的財産権に関する EU 法で規定されている、知的財産に関する基本的権利[8] および権利の行使における効果的な救済に関する基本的権利の行使を促進するために必要である。著作権および関連権利の場合、訓練データの透明性は、汎用 AI モデル提供者が著作権および関連権利に関する EU 法を遵守することを確保するのに貢献する[9]。 |
| (9) Second, transparency of the training data in the Summary may facilitate data subjects’ rights and more broadly support the enforcement of the Union data protection rules. In particular, this can be done by summarising all the relevant information together, such as information about the data scraped from the internet or collected by the provider through interactions with the model or other services and products. The information in the Summary is not meant to replace, nor affect the respective information the providers of general-purpose AI models should make available to data subjects under Union data protection law. In the context of the Summary, the interests of consumers and the protection of their consumer rights under Union law may also be relevant. |
(9) 第二に、概要における訓練データの透明性は、データ対象者の権利の行使を促進し、より広く EU のデータ保護規則の執行を支援する可能性がある。これは、インターネットから収集したデータ、または提供者がモデルやその他のサービスおよび製品との相互作用を通じて収集したデータに関する情報など、関連するすべての情報をまとめて要約することで実現できる。概要に記載される情報は、汎用 AI モデル提供者が EU データ保護法に基づきデータ対象者に提供すべき情報を置き換えるものではなく、また、その情報に影響を与えるものでもない。概要に関しては、EU 法に基づく消費者の利益および消費者権利の保護も関連する場合がある。 |
| (10) Third, transparency of the general characteristics of the content used for training may also assist providers integrating these models into downstream applications to assess the diversity of the data. This, in turn, will allow them to implement, where appropriate, mitigating measures to ensure that the fundamental rights to non-discrimination[10] and language and cultural diversity[11] are respected. |
(10) 第三に、訓練に使用されるコンテンツの一般的な特性の透明性は、これらのモデルをダウンストリームアプリケーションに統合する提供者が、データの多様性を評価する上でも役立つ可能性がある。これにより、提供者は、必要に応じて、非差別[10] および言語および文化の多様性[11] に関する基本的権利が尊重されるように、緩和措置を実施することができる。 |
| (11) Fourth, greater transparency of the training data may also facilitate the fundamental right to receive and impart information[12] and allow researchers to exercise their freedom of science[13] to conduct scientific research. It can allow academic institutions and organisations to critically evaluate the implications and limitations of a particuler general-purpose AI model and the potential risks and harms associated with the data used. |
(11) 第四に、訓練データの透明性を高めることで、情報を受け取り、伝達する基本的権利[12] の行使が促進され、研究者が科学的研究を行うための科学的自由[13] の行使が可能になる。これにより、学術機関や学術団体は、特定の汎用 AI モデルの影響や限界、および使用されるデータに関連する潜在的なリスクや危害を批判的に評価することができる。 |
| (12) Finally, transparency of the training data may also contribute to more transparent and competitive markets. For example, information about whether publicly available general-purpose AI models have been used to train other models, in particular through model distillation, or whether a model has been trained on user data collected from provider’s own products and services, may help users and companies better understand how their data and models have been used and avoid potential lock-in effects. |
(12) 最後に、訓練データの透明性は、より透明で競争力のある市場にも貢献する可能性がある。例えば、公開されている汎用 AI モデルが、特にモデル蒸留によって他のモデルの訓練に使用されているかどうか、あるいはモデルが提供者の自社製品やサービスから収集したユーザーデータを用いて訓練されているかどうかに関する情報は、ユーザーや企業が自社のデータやモデルがどのように使用されているかをよりよく理解し、潜在的なロックイン効果を回避するのに役立つ。 |
| 3. Comprehensive scope of the training data and sufficient details |
3. 訓練データの包括的な範囲と十分な詳細 |
| (13) Information about the general-purpose AI model provided in the Summary should cover data used in all stages of the model training, from pre-training to post-training, including model alignment and finetuning. This covers all sources and types of data, regardless of whether the data are protected or not, including by an intellectual property right. Since Article 53(1)(d) AI Act mentions explicitly ‘training’, other input data used during the model’s operation (e.g. through retrieval augmented generation) are not required in the mandatory sections of the Template[14], unless the model actively learns from this input data. |
(13) 概要で提供される汎用 AI モデルに関する情報は、事前学習から事後学習、モデルの調整や微調整に至るまで、モデルの訓練のすべての段階で使用されるデータについて網羅すべきだ。これには、知的財産権によって保護されているかどうかに関わらず、すべてのデータ源およびデータの種類が含まれる。AI 法第 53 条(1)(d) では「訓練」が明示的に記載されているため、モデルの運用中に使用されるその他の入力データ(検索拡張生成によるものなど)は、モデルが当該入力データから積極的に学習する場合を除き、テンプレートの必須項目には記載する必要はない[14]。 |
| (14) Recital 107 AI Act explains that the information about the training content should be comprehensive in its scope and sufficiently detailed to achieve the objective of the Summary of providing meaningful public transparency and facilitating parties with legitimate interests to exercise and enforce their rights under Union law. |
(14) AI 法前文 107 では、訓練の内容に関する情報は、要約の目的である、意味のある公開の透明性を確保し、正当な利益を有する当事者が EU 法に基づく権利を行使および執行することを容易にするために、その範囲が包括的かつ十分に詳細であるべきであると説明されている。 |
| (15) The Template annexed to this Explanatory Notice aims to provide a common minimal baseline for the information to be made publicly available in the Summary. It consists of three main sections: |
(15) この説明通知に添付のテンプレートは、概要で公開すべき情報の共通最低基準を提供することを目的としている。テンプレートは、主に 3 つのセクションで構成されている。 |
| 1. General information: this section requires information allowing identification of the provider and of the model, and information on modalities, the size of each modality within broad ranges, as well as general characteristics of the training data. |
1. 一般的な情報:このセクションでは、提供者およびモデルを特定するための情報、モダリティ、広範な範囲における各モダリティの規模、および訓練データの一般的な特性に関する情報を記載する必要がある。 |
| 2. List of data sources: this section requires disclosure of the main datasets that were used to train the model, such as large private or public databases, and a comprehensive narrative description of the data scraped online by or on behalf of the provider (including a summary of the most relevant domain names scraped) and a narrative description of all other data sources used (e.g. user data or synthetic data) to ensure completeness of the summary regarding the content used for the model training[15]. |
2. データソースのリスト:このセクションでは、モデルの訓練に使用された主なデータセット(大規模な民間または公的データベースなど)の開示、および提供者またはその代理人がオンラインで収集したデータに関する包括的な説明(最も関連性の高いドメイン名の概要を含む)と、モデルの訓練に使用されたコンテンツに関する要約の完全性を確保するために使用されたその他のすべてのデータソース(ユーザーデータや合成データなど)に関する説明の開示が義務付けられている。[15]。 |
| 3. Relevant data processing aspects: this section of the Template requires disclosure of certain data processing aspects that are relevant for the exercise of the rights of parties with legitimate interests under Union law. This is especially important for compliance with Union law on copyright and related rights and for the removal of illegal content to mitigate the risk that such illegal content may be reproduced and disseminated at scale by the general-purpose AI model. |
3. 関連するデータ処理の側面:テンプレートのこのセクションでは、EU 法に基づく正当な利益を有する当事者の権利の行使に関連する特定のデータ処理の側面を開示する必要がある。これは、著作権および関連権利に関する EU 法の遵守、および汎用 AI モデルによって違法なコンテンツが大規模に複製および拡散されるリスクを緩和するための違法コンテンツの削除にとって特に重要である。 |
| (16) On the basis that the Summary aims to provide sufficient details and to facilitate parties with legitimate interests, including rightsholders, exercising their rights under Union law, the Template requires a disclosure of a summary of the list of top domain names crawled and scraped from online sources in a summarised narrative form[16]. At the same time, it does not require disclosure of the details for the specific data and works used to train the model as this would go beyond the requirement in Article 53(1)(d) to provide just a ‘summary’, which in line with Recital 107 AI Act must be “generally comprehensive” but not “technically detailed”. Providers may nevertheless decide on a voluntary basis to go beyond the minimum requirements in the Template and disclose in the Summary more details than what is required by Article 53(1)(d) AI Act and the Template[17]. Furthermore, for domain names scraped or crawled from the internet that are not listed in the Summary, it is recommended that providers act in good faith and on a voluntary basis enable parties with a legitimate interest including rightholders, upon requests, to obtain information whether the provider has scraped and used for training content which includes protected works and other subject matter that rightholders have made available on specific internet domains. This recommended and voluntary ‘upon request’ mechanism does not affect other available remedies for rightholders under Union law on enforcement of intellectual property rights (e.g. Article 8 of the Intellectual Property Rights Enforcement Directive)[18]. |
(16) 概要は、十分な詳細情報を提供し、権利者を含む正当な利益を有する当事者が EU 法に基づく権利を行使することを容易にするものであることを踏まえ、テンプレートでは、オンラインソースからクロールおよびスクレイピングしたトップドメイン名のリストの概要を、要約した記述形式で[16]開示することを義務付けています。同時に、モデルを訓練するために使用された特定のデータおよび著作物の詳細を開示することは、AI 法前文 107 項に従い、「概要」のみを提供するという第 53 条(1)(d) の要件を超えるため、テンプレートでは開示を義務付けていない。ただし、提供者は、テンプレートの最低要件を超えて、AI 法第 53 条(1)(d) およびテンプレートで要求される情報よりも詳細な情報を概要で開示することを自主的に決定することができる[17]。さらに、要約に記載されていない、インターネットからスクレイピングまたはクロールされたドメイン名については、提供者は、権利者を含む正当な利益を有する当事者からの要請に応じて、提供者が、権利者が特定のインターネットドメインで公開した保護対象作品およびその他の対象物を含むコンテンツをスクレイピングし、訓練用に使用したかどうかに関する情報を、誠意をもって、かつ自主的に提供することが推奨される。この推奨される任意的な「要請に応じて」の仕組みは、知的財産権の執行に関する連合法(例えば、知的財産権執行指令の第8条)[18]に基づく権利者の他の利用可能な救済措置に影響を与えない。 |
| 4. Balance with trade secrets and confidential business information |
4. 営業秘密および機密の事業情報とのバランス |
| (17) As explained in Recital 107 AI Act, the Template should seek to strike a balance between serving the interests of parties with legitimate interests and promoting increased transparency of the training content in a meaningful way, while respecting the rights of all parties concerned, in particular taking due account of the need to protect trade secrets and confidential business information. |
(17) AI 法前文 107 で説明されているように、テンプレートは、正当な利益を有する当事者の利益と、訓練コンテンツの透明性の向上とのバランスを、すべての関係者の権利を尊重しつつ、特に営業秘密および機密の事業情報を保護する必要性を十分に考慮して、有意義な形で図るべきである。 |
| (18) Since the Commission is bound by the Charter on fundamental rights, this careful balancing exercise has been implemented in relation to the information that the Template requires to be disclosed in order for providers to fulfil their obligation under Article 53(1)(d) AI Act and provide a ‘sufficiently detailed’ public summary of the training content. The provision of information regarding more specific details about the content used for the training of the general-purpose AI models is required in the Template only where it is necessary to enable the exercise of rights protected under Union law in a meaningful manner as required by Article 53(1)(d) and Recital 107 AI Act. Determining which details should be disclosed has been the result of a careful balancing exercise carried out by the Commission in drawing up the Template to ensure that relevant information on the training data is provided to meet the Template’s objectives, while confidential commercially sensitive information about the data sources and the precise manner in which providers curate the data and train their models is preserved. |
(18) 委員会は基本権憲章に拘束されているため、テンプレートが、提供者が AI 法第 53 条(1)(d) の義務を履行し、訓練内容について「十分に詳細な」公開要約を提供するために開示を義務付けている情報について、この慎重なバランス調整を実施している。汎用 AI モデルの訓練に使用される内容に関するより具体的な詳細情報の提供は、AI 法第 53 条(1)(d) および前文 107 で要求される、EU 法で保護される権利を意味のある形で行使するために必要な場合にのみ、テンプレートで要求されている。開示すべき詳細を決定するには、テンプレートの作成において、テンプレートの目的を達成するために訓練データに関する関連情報を提供すると同時に、データソースや提供者がデータをキュレーションし、モデルを訓練する正確な方法に関する商業的に機密性の高い情報を保護するために、委員会が慎重なバランス調整を行った結果によるもの。 |
| (19) To protect providers’ trade secrets, different levels of detail are required in the Template depending on the source of data considered. In particular, limited disclosure of information is required for licensed data given that the rightsholders concerned are parties to the licensing agreements (see Section 2.2.1 of the Template). Furthermore, private datasets not commercially licensed by rightsholders and obtained from other third parties have to be listed only if publicly known (or the provider wants to make them publicly known), and otherwise described in a general manner (see Section 2.2.2 of the Template). Considering the public nature of the information contained in publicly available datasets, more detail is required about those datasets, including the disclosure of ‘large’ datasets (defined in the Template), in line with Recital 107 AI Act (see Section 2.1 of the Template). |
(19) 提供者の営業秘密を保護するため、テンプレートでは、検討対象のデータ源に応じて、詳細度の異なる情報開示が求められている。特に、ライセンス供与されたデータについては、権利者がライセンス契約当事者であることを考慮し、情報開示を制限している(テンプレートの 2.2.1 項を参照)。さらに、権利者によって商業的にライセンス供与されておらず、他のサードパーティから取得した非公開のデータセットは、公に知られている場合(または提供者が公に知らしめたい場合)にのみ記載し、それ以外の場合は一般的な方法で記述する必要がある(テンプレートの 2.2.2 項を参照)。公開データセットに含まれる情報の公共性を考慮し、AI 法前文 107 項(テンプレート 2.1 参照)に従い、テンプレートで定義される「大規模」データセットの開示を含め、これらのデータセットについてより詳細な情報が必要とされる。 |
| (20) For data scraped from online sources, the Template requires disclosure of relevant information such as the crawlers used, their purpose and behaviour, the period of collection and a comprehensive description of the type of content and online sources scraped (see Section 2.3 of the Template). In addition, the Template requires disclosure of a summary list of most relevant domain names crawled and scraped from online sources by or on behalf of the provider in a summarized narrative form[19], in so far as their content has been used for the training of the general-purpose AI model. Such a summary of the domain names scraped from the internet aims to provide a meaningful information about the most relevant top domain names scraped, while striking a balance with the trade secrets ensuring the Summary remains non-technical and in a summarised narrative form, as required by Recital 107 AI Act. |
(20) オンラインソースからスクレイピングされたデータについては、テンプレートは、使用されたクローラー、その目的と動作、収集期間、およびスクレイピングされたコンテンツの種類とオンラインソースの包括的な説明を含む関連情報の開示を要求している(テンプレートの第2.3節参照)。さらに、テンプレートでは、その内容が汎用 AI モデルの訓練に使用されている場合、提供者またはその代理人がオンラインソースからクロールおよびスクレイピングした最も関連性の高いドメイン名の要約リストを、要約した記述形式[19]で開示することを義務付けています。インターネットからスクレイピングしたドメイン名のこのような要約は、スクレイピングした最も関連性の高いトップドメイン名に関する有意義な情報を提供すると同時に、前文 107 AI 法で要求されているように、要約が技術的ではなく、要約された記述形式であることを確保し、営業秘密とのバランスを保つことを目的としている。 |
| (21) Furthermore, the Template (see Section 2.4) requires minimal information about user data collected through user interactions with all services and products of the provider, including interactions with the providers’ AI models. This category excludes data licensed by users based on commercial transactional agreements already covered under Section 2.2.1, or customer data used for fine-tuning a model for specific purposes. For synthetic data generated by AI model(s) used for training purposes and in particular for model distillation (see Section 2.5 of the Template), the information is also limited to names of the general-purpose AI model(s) used if those models have been placed on the market or, if other AI models have been used, including models owned by the providers, information about the model (including a general description of the model training data if known and in so far as this may be needed for the exercise of the rights of parties with legitimate inetersts and to avoid circumvention of the disclosure obligations in the other Sections of the Template). |
(21) さらに、テンプレート(2.4 節を参照)では、提供者の AI モデルとのやり取りを含め、提供者のすべてのサービスおよび製品とのユーザーのやり取りを通じて収集されるユーザーデータに関する最小限の情報のみを要求している。このカテゴリーには、2.2.1 節で既にカバーされている商業取引契約に基づいてユーザーからライセンス供与されたデータ、または特定の目的のためのモデルの微調整に使用される顧客データは含まれない。訓練目的、特にモデル蒸留(テンプレートセクション 2.5 を参照)に使用される AI モデルによって生成された合成データについては、そのモデルが上市されている場合は、使用された汎用 AI モデルの名称、または提供者が所有するモデルを含む他の AI モデルが使用されている場合は、そのモデルに関する情報(モデル訓練データに関する一般的な説明(既知の場合、および正当な利益を有する当事者の権利の行使のために必要であり、テンプレートの他のセクションにおける開示義務の回避を回避するために必要な場合)を含む)に限定される。(モデル訓練データの一般的な説明を含む。ただし、正当な利益を有する当事者の権利の行使や、テンプレートの他のセクションにおける開示義務の回避を避けるために必要とされる範囲に限定する)。 |
| (22) The Template does not require disclosure of the exact mix and composition of data sources, but only high-level information about the training data size per modality (selection amongst three very broad ranges) and aggregated across all sources[20] (see Section 1.2 of the Template). |
(22) テンプレートでは、データソースの正確な組み合わせや構成の開示は求められていないが、モダリティごとの訓練データサイズ(3つの非常に広範な範囲から選択)に関する高水準の情報と、すべてのソースを統合した情報[20](テンプレートの第1.2節を参照)の開示が求められている。 |
| 5. Simple, uniform and effective reporting |
5. シンプルで統一された効果的な報告 |
| (23) The information requested by the Template is to be provided in a narrative, simple and effective form. The Template aims to ensure the reported information is useful and understandable to the public and to the parties concerned, while avoiding unnecessary burden on providers of general-purpose AI models, including SMEs. |
(23) テンプレートで要求される情報は、説明的、簡潔かつ効果的な形で提供しなければならない。テンプレートは、中小企業を含む汎用 AI モデル提供者に不必要な負担をかけることを避けながら、報告された情報が一般市民および関係者に有用で理解しやすいものとなることを目指している。 |
| (24) Each Section of the Template includes clear and short instructions to allow providers to report the required information in an easy and uniform manner. The Commission aims to provide the Template as an online form and to publish it on its website. |
(24) テンプレートの各セクションには、提供者が要求される情報を簡単かつ統一的な方法で報告できるよう、明確かつ簡潔な指示が記載されている。欧州委員会は、テンプレートをオンラインフォームとして提供し、そのウェブサイトに掲載することを目指している。 |
| (25) Providers should ensure that the information included in the Summary is reported in good faith and in an accurate and comprehensive manner. Flexibility is provided under specific sections, as indicated in the Template, to disclose only information that is relevant, necessary for the purpose of the Summary, and practicable to obtain (e.g. regarding the categorisation of some of the content or the characteristics of the training data, or the period of data collection). |
(25) 提供者は、要約に含まれる情報が、誠実かつ正確かつ包括的に報告されていることを確保すべきだ。テンプレートに示されているように、要約の目的に関連し、必要かつ入手可能な情報(一部のコンテンツの分類、訓練データの特性、データ収集期間など)のみを開示するよう、特定のセクションでは柔軟性が認められている。 |
| (26) The AI Office may verify whether the Template has been filled in correctly in order to assess if the provider has complied with Article 53(1)(d) AI Act. In this context, the AI Office has all enforcement powers under the AI Act and and may request corrective measures. Non-compliance may be sanctioned with fines of up to 3% of the provider’s annual total worldwide turnover in the preceding financial year or EUR 15 000 000, whichever is higher. The lawful collection and processing of the data remains the responsibility of the provider under other applicable Union law (e.g. copyright and data protection). The AI Office will supervise the implementation of the obligation to provide a compliant summary under Article 53(1)(d) AI Act without performing a work-by-work assessment or checks whether specific content has been used or not for the training of the general-purpose AI model (Recital 108 AI Act). |
(26) AI 事務局は、提供者が AI 法第 53 条(1)(d) を遵守しているかどうかを評価するために、テンプレートが正しく記入されているかどうかを検証することができる。この文脈において、AI 事務局は AI 法に基づくすべての執行権限を有し、是正措置を要求することができる。違反した場合は、提供者の直前会計年度の全世界の年間総売上高の 3% または 15,000,000 ユーロのうち、いずれか高い方の罰金が科せられる。データの合法的な収集および処理は、その他の適用される EU 法(著作権法、データ保護法など)に基づき、提供者の責任とする。AI 事務局は、AI 法第 53 条(1)(d)に基づく準拠した要約の提供義務の履行を監督するが、作品ごとの評価や、汎用 AI モデルの訓練に特定のコンテンツが使用されたかどうかの確認は行わない(AI 法前文 108)。 |
| (27) In case of disputes, providers and parties with a legitimate interest, including rightsholders, are encouraged to use alternative dispute resolution mechanisms available at national level (such as mediation) and other available remedies provided for by Union and national law (e.g. under Article 8 of the Intellectual Property Rights Enforcement Directive)[21]. |
(27) 紛争が生じた場合、提供者および権利者を含む正当な利益を有する当事者は、国内レベルで利用可能な代替的紛争解決メカニズム(調停など)および EU 法および国内法(知的財産権執行指令第 8 条など)で規定されているその他の救済手段を利用することが奨励される[21]。 |
| 6. Modifications of existing general-purpose AI models and updates |
6. 既存の汎用 AI モデルの修正および更新 |
| (28) An existing general-purpose AI model already placed on the Union market may be modified by a downstream entity in such a way that the downstream entity becomes the provider of the resulting general-purpose AI model, as specified in the Commission guidelines on General-Purpose AI models[22]. In such cases, the information reported by the modifying entity in the Template should be limited to the training content used for the model modification only[23], and the name of the model(s) that was modified should be clearly indicated in the Summary (see Section 1.2 of the Template). |
(28) EU 市場にすでに上市されている既存の汎用 AI モデルは、欧州委員会による汎用 AI モデルに関するガイドライン[22] に規定されているように、下流事業体によって、その下流事業体が結果としての汎用 AI モデル提供者となるような方法で変更することができる。そのような場合、変更事業体がテンプレートで報告する情報は、モデル変更に使用された訓練内容のみに限定すべきであり[23]、変更されたモデル名は要約に明記すべきである(テンプレートの 1.2 節を参照)。 |
| (29) The Summary should also be updated whenever the provider further trains its own general-purpose AI model placed on the market on additional data that requires an update of the content of the Summary. In those cases, the Summary should be updated at six-month intervanls or if in the meantime the additional data used to further train the model requires a materially significant update of the content of the Summary, whichever is sooner. In such cases, the Summary should be updated to reflect this additional data, as well as the date of the update. The updated Summary should be made publicly available alongside the modified model. |
(29) 提供者が、上市した自社の汎用 AI モデルを、要約の内容の更新を必要とする追加データを用いてさらに訓練した場合も、要約を更新すべきだ。このような場合、要約は6ヶ月ごとに更新するか、またはその間にモデルをさらに訓練するために使用された追加データが要約の内容に重大な変更を要する場合、いずれか早い方の日付で更新する必要がある。このような場合、要約は追加データを反映し、更新日を表示するように更新する必要がある。更新された要約は、変更されたモデルと共に公開する必要がある。 |
| (30) The same Summary may be used for different models or different model versions if the content of their respective Summaries is identical. In this case, the Summary should clearly specify the different models and model version to which it applies. In addition, if different models or model versions are based on the same general-purpose AI model that has already been placed on the Union market, and the Summaries for each model and model version are different so that they cannot be covered by a single Summary, the Summaries for each of those models or model versions only need to cover the training data specifically used to further modify (including fine-tune) them out of the original model. In this case, a clear reference should be made to the original model in the Summary for each relevant model or model version, and a link to the Summary of the original model included (see Section 1.2). |
(30) 異なるモデルまたはモデルバージョンに対して同じ要約を使用することができる。ただし、それぞれの要約の内容が同一である場合に限る。この場合、要約は、適用される異なるモデルおよびモデルバージョンを明確に指定しなければならない。さらに、異なるモデルまたはモデルバージョンが、すでに EU 市場に上市されている同じ汎用 AI モデルに基づいており、各モデルおよびモデルバージョンの概要が 1 つの概要では網羅できないほど異なる場合、各モデルまたはモデルバージョンの概要は、元のモデルからさらに変更(微調整を含む)するために具体的に使用された訓練データのみを対象とすれば十分である。この場合、各関連するモデルまたはモデルバージョンの要約において、元のモデルへの明確な参照を記載し、元のモデルの要約へのリンクを含めること(第1.2項参照)。 |
| (31) Where the same Summary is used for multiple models or model versions, in accordance with point (30), reference in the template to the ‘model’ should be understood as a reference to each model or model version covered by the Summary. Reference to ‘training data’ should be understood as a reference to the training data for each of these models or model versions. |
(31) 同じ要約が複数のモデルまたはモデルバージョンに適用される場合、点 (30) に従い、テンプレート内の「モデル」への参照は、要約でカバーされる各モデルまたはモデルバージョンへの参照と解釈される。 「訓練データ」への参照は、これらの各モデルまたはモデルバージョン用の訓練データへの参照と解釈される。 |
| 7. Publication of the Summary |
7. 要約の公表 |
| (32) The Summary should be made publicly available at the latest when the model is placed on the Union market. It should be published on the provider’s official website in a clearly visible and accessible manner, clearly indicating which model(s) (and possibly model version(s)) the Summary covers subject to the conditions specified in point (30) above. The Summary should also be made publicly available together with the model across all its public distribution channels (e.g. online platforms). |
(32) 要約は、遅くともモデルが EU 市場に上市された時点で、一般に公開されるべきだ。要約は、提供者の公式ウェブサイトに、明確かつ目立ち、アクセスしやすい形で公表され、上記 (30) で規定された条件に従って、要約が対象とするモデル(およびモデルバージョンがある場合はそのモデルバージョン)を明確に示すべきだ。要約は、モデルと共に、すべての公開配布チャネル(例:オンラインプラットフォーム)を通じて公開する必要がある。 |
| 8. Entry into application of the obligation and special rules for models placed on the market before 2 August 2025 |
8. 2025 年 8 月 2 日より前に上市されたモデルに対する義務の適用開始および特別規則 |
| (33) The obligation for making the Summaries publicly available becomes applicable as of 2 August 2025. For models placed on the market before 2 August 2025, providers should take the necessary steps to make the corresponding Summary publicly available no later than 2 August 2027. Where a provider of a model placed on the market before 2 August 2025 cannot, despite their best efforts, provide parts of the information required to prepare the Summary because the information is not available or its retrieval would impose a disproportionate burden on the provider, the provider should clearly state and justify the corresponding information gaps in its Summary[24]. The supervision and enforcement by the AI Office for compliance with the rules for general-purpose AI models will start as of 2 August 2026. |
(33) 概要を一般に公開する義務は、2025 年 8 月 2 日から適用される。2025 年 8 月 2 日より前に上市されたモデルについては、提供者は、2027 年 8 月 2 日までに、対応する概要を一般に公開するために必要な措置を講じるべきだ。2025 年 8 月 2 日より前に上市されたモデルの提供者が、最善の努力にもかかわらず、要約の作成に必要な情報の一部を提供できない場合(その情報が利用できない、またはその情報の取得に提供者に過度の負担がかかる場合)、提供者は、要約において、その情報不足を明記し、その理由を明らかにすべきだ[24]。AI事務局による汎用AIモデルに関する規則の遵守状況の監督および執行は、2026年8月2日から開始される。 |
| 9. Review of the Explanatory Notice and the Template |
9. 説明通知およびテンプレートの見直し |
| (34) The Commission will monitor the implementation of the Template annexed to this Explanatory Notice and where necessary review the Notice and the Template, in view of practical experience gained and of the pace of technological, societal and market developments in this area. If the Commission deems it necessary, such a review may take place before the entry into application of the enforcement powers of the AI Office on 2 August 2026. |
(34) 委員会は、この説明通知に付属のテンプレートの実施状況を監視し、この分野における実務経験や技術、社会、市場の進展のペースを踏まえて、必要に応じて、この通知およびテンプレートを見直す。委員会が必要と認めた場合、このような見直しは、2026 年 8 月 2 日に AI 事務局の執行権限が適用される前に実施されることがある。 |
| |
|
| [1] Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act) (Text with EEA relevance), PE/24/2024/REV/1, OJ L, 2024/1689, 12.7.2024. |
[1] 欧州議会および理事会の規則(EU)2024/1689(2024年6月13日)は、人工知能に関する調和された規則を定めるとともに、規則(EC)第300/2008号、規則(EU)第167/2013号、 (EU)第168/2013号、(EU)第2018/858号、(EU)第2018/1139号、(EU)第2019/2144号、および指令2014/90/EU、(EU)第2016/797号、(EU)第2020/1828号を改正するもの (人工知能法)(EEA関連テキスト)、PE/24/2024/REV/1、OJ L、2024/1689、2024年7月12日。 |
| [2] Article 53 AI Act. |
[2] AI法第53条。 |
| [3] Article 55 AI Act |
[3] AI法第55条 |
| [4] The exception for free and open-source general-purpose AI model under Article 53(2) AI Act does not apply to the obligation to make publicly available the Summary. |
[4] AI 法第 53 条(2) の、自由でオープンソースの汎用 AI モデルに関する例外は、概要の公開義務には適用されない。 |
| [5] See Article 2 AI Act and Guidelines for providers of general-purpose AI models | Shaping Europe’s digital future |
[5] AI 法第 2 条および汎用 AI モデル提供者向けガイドライン | ヨーロッパのデジタルの未来を形作る |
| [6] AI Act: Have Your Say on Trustworthy General-Purpose AI | Shaping Europe’s digital future |
[6] AI 法:信頼性の高い汎用 AI についてのご意見をお聞かせください | ヨーロッパのデジタルの未来を形作る |
| [7] General-Purpose AI Code of Practice | Shaping Europe’s digital future |
[7] 汎用AI行動規範 | ヨーロッパのデジタルの未来を形作る |
| [8] Article 17(2) of the EU Charter of Fundamental Rights of the European Union, OJ C 326, 26.10.2012, p. 391–407. |
[8] 欧州連合基本権憲章第17条(2)、OJ C 326、2012年10月26日、391–407ページ。 |
| [9] Directive (EU) 2019/790 of the European Parliament and of the Council of 17 April 2019 on copyright and related rights in the Digital Single Market and amending Directives 96/9/EC and 2001/29/EC (Text with EEA relevance.), PE/51/2019/REV/1, OJ L 130, 17.5.2019, p. 92–125. |
[9] 欧州議会および理事会指令(EU)2019/790(2019年4月17日)デジタル単一市場における著作権および関連権利に関する指令、および指令96/9/ECおよび2001/29/ECの改正(EEA関連テキスト) PE/51/2019/REV/1、OJ L 130、2019年5月17日、92–125頁。 |
| [10] Article 21 of the EU Charter of Fundamental Rights. |
[10] 欧州連合基本権憲章第21条。 |
| [11] Article 22 of the EU Charter of Fundamental Rights. |
[11] 欧州連合基本権憲章第22条。 |
| [12] Article 11(1) of the EU Charter of Fundamental Rights. 13 |
[12] 欧州連合基本権憲章第11条第1項。13 |
| [13] Article 13 of the EU Charter of Fundamental Rights. |
[13] 欧州連合基本権憲章第13条。 |
| [14] Since such data is used as input for the generation process, its influence on the outputs of the model may be significant, and relevant for the exercise of the rights of parties with legitimate interests. For this reason it may be disclosed by the provider on a voluntary basis in the optional Sections in the Template which allow the provision of additional information. |
[14] このようなデータは生成プロセスの入力として使用されるため、モデルの出力に大きな影響を与える可能性があり、正当な利益を有する当事者の権利の行使に関連する場合がある。このため、提供者は、追加情報の提供を認めるテンプレートの任意欄に、自主的に開示することができる。 |
| [15] See in this context Recital 107 of the AI Act. |
[15] この点については、AI 法の前文 107 項を参照のこと。 |
| [16] See Section 2.3. of the Template that requires a list of the internet domain names (top and second-level domain, e.g. “example.com”) in the top 10 % of all domain names determined by the size of the content scraped (in a representative manner across modalities, where applicable). For small and medium-sized enterprises (SMEs), including start-ups, the Template requires the internet domain names in the top 5%, or the top 1000 domains to ensure proportionality of the burden on SMEs in line with recital 109 AI Act. |
[16] テンプレートのセクション 2.3 を参照。このセクションでは、コンテンツのスクレイピングの規模(該当する場合、モダリティ全体で代表的な方法で)によって決定される、すべてのドメイン名のトップ 10% に属するインターネットドメイン名(トップレベルおよびセカンドレベルドメイン、例:「example.com」)のリストの提出が義務付けられている。スタートアップを含む中小企業(SME)については、AI 法前文 109 に基づき、SME への負担の均衡を確保するため、テンプレートでは、上位 5% のインターネットドメイン名、または上位 1000 ドメインの記載が義務付けられている。 |
| [17] See in the end of each sub-section of the Template, optional part with possibilities to provide additional information. |
[17] テンプレートの各サブセクションの末尾にある、追加情報を提供できる任意の部分を参照のこと。 |
| [18] Directive 2004/48/EC of the European Parliament and of the Council of 29 April 2004 on the enforcement of intellectual property rights (OJ L 157, 30.4.2004), OJ L 195, 2.6.2004, p. 16–25. |
[18] 欧州議会および理事会指令2004/48/EC(2004年4月29日)知的財産権の執行に関する指令(OJ L 157, 30.4.2004)、OJ L 195, 2.6.2004, p. 16–25。 |
| [19] See footnote 16 above and Section 2.3. of the Template. |
[19] 上記脚注16およびテンプレートの2.3節を参照。 |
| [20] Ref to ECJ case-law aligned with this approach. |
[20] このアプローチと一致する欧州司法裁判所(ECJ)の判例を参照。 |
| [21] Directive 2004/48/EC of the European Parliament and of the Council of 29 April 2004 on the enforcement of intellectual property rights (OJ L 157, 30.4.2004), OJ L 195, 2.6.2004, p. 16–25. |
[21] 欧州議会および理事会指令2004/48/EC(2004年4月29日)知的財産権の執行に関する指令(OJ L 157、2004年4月30日)、OJ L 195、2004年6月2日、16–25頁。 |
| [22] Guidelines for providers of general-purpose AI models | Shaping Europe’s digital future |
[22]汎用 AI モデル提供者向けガイドライン | ヨーロッパのデジタルの未来を形作る |
| [23] See also Recital 107 AI Act. |
[23] AI 法前文 107 も参照のこと。 |
| [24] See under each Section of the Template a box for possible additional information (optional). |
[24] テンプレートの各セクションの下にある、追加情報(任意)を入力する欄を参照のこと。 |
Recent Comments