Whitepaper: Proposal to leverage Article 17 to build a public repository of Public Domain and openly licensed works.
Article 17 is by far the most controversial article of the 2019 Copyright in the Digital Single Market directive. This article establishes a new liability regime for so-called Online Content Sharing Service Providers (OCSSPs) operating in the EU. The final version of Article 17, which is the outcome of long and contentious negotiations, has a number of sometimes conflicting objectives within its 10 sub-paragraphs. Among these is the obligation on platforms to cooperate with rightholders to make best efforts to prevent the availability of works that rightholders want to be blocked in Article 17(4) and the requirement – in Article 17(7) – that such cooperations between rightholders and platforms do not result in the prevention of the availability of works uploaded by users, which do not infringe copyright.
Reconciling these two seemingly contradictory requirements is one of the key challenges facing Member States in their national implementations of the directive – and by extension for OCSSPs who will need to ensure that their practices comply with them. And while the fundamental rights compliance of the provisions contained in Article 17(4) is still under review by the CJEU1, the first Member States have already enacted national implementations. The most notable attempt to reconcile the conflicting obligations in practice can be found in the German implementation law that has come into effect on the 1st of August 20212.
The German implementation includes a number of provisions that are specifically designed to reduce the risk of so-called overblocking: The unjustified blocking or removal of uploads subsequent to rightholder requests to prevent the availability of their works in accordance with Article 17(4) of the directive. These provisions include the requirement not to block “presumably legitimate” uploads and the requirement to keep disputed uploads available until the dispute is resolved.
Measures to prevent openly licensed and Public Domain works from being blocked
In addition the German implementation law contains a specific provision aimed at preventing the unjustified blocking of works that are in the Public Domain or that have been licensed under open licenses. Section 18(4) UrhDaG states:
After an abusive blocking request in respect of works in the public domain or works whose use by anyone is authorised free of charge, service providers must ensure, to the best of their ability and in accordance with section 1 (2), that these works are not blocked again3.
In this context works "whose use by anyone is authorised free of charge" must be understood to include works that have been licensed under a Creative Commons license or another open content license.
To comply with this provision, OCSSPs operating in Germany will need to maintain an internal repository of works for which they have (1) received a blocking request and where (2) such blocking request has turned out to be abusive because the works are either in the Public Domain or where the use of the work is authorised under the terms of an open license. The abusive nature of a blocking request can either be identified by the OCSSP by itself or – more likely – become known to the OCSSP as the result of a user challenging a block of uploaded content through the complaint and redress mechanism specified in Article 17(9) of the Directive (as implemented in Articles 14 to 17 UrhDaG).
Over time such repositories can be expected to grow and will likely start to contain a substantial number of entries relating to a wide variety of openly licensed and public domain works. This will result in the repositories obtaining value beyond the relatively narrow use case of preventing overblocking of openly licensed and PD works by OCSSPs: They will become repositories of freely reusable works that can help to unlock the societal value of these works4. For entities falling within the scope of the German UhrDaG maintaining such a database is necessary to comply with the provisions of the law. But even outside of this relatively narrow and well defined scope, having access to such a database will help OCSSPs to comply with the provisions of the DSM directive. In this context it is worth recalling that Article 17(7) of the DSM Directive contains the requirement that:
The cooperation between online content-sharing service providers and rightholders shall not result in the prevention of the availability of works or other subject matter uploaded by users, which do not infringe copyright and related rights, including where such works or other subject matter are covered by an exception or limitation.
Legitimate uses which do not infringe copyright or related rights may include (a) uses under exceptions and limitations, (b) uses by those who hold or have cleared the rights in the content they upload or uses covered by the authorisation under Article 17(2) and (c) uses of content not covered by copyright or related rights, notably works in the public domain or, for example, content where the threshold of originality or any other requirement pertaining to the threshold for protection, is not met.
As a result - and within the limits imposed by the proportionality principle from Article 17(5) - all platforms qualifying as OCSSPs are required to award special protection from automated blocking systems to openly licensed works and works in the Public Domain.
While the obligation in Article 17(7) CDSM (and the obligation in Section 18(4) UhrDaG which is derived from it) are obligations imposed on each OCSSP individually, it is possible for OCSSPs to share such repositories with other OCSSPs (and with platforms who fall outside of the scope of the Article 17, but who function in a similar fashion to these platforms) in order to pool information on openly licensed works and works in the public domain. If executed well, pooling this information could both increase the quality and quantity of the information (thereby improving compliance with the general principle established in Article 17(7)) and to reduce the costs of compliance (assuming that operating a shared database requires fewer resources than maintaining platform- or operator-specific repositories).
The most important advantage of such pooling of information would reside in the fact that it has the potential to prevent unjustified blockings that have occurred on one platform from happening on another platform making use of the same shared repository. Having access to pooled information on openly licensed and Public Domain works would be especially beneficial for (the users of) smaller platforms as they would have access to larger libraries of works than they can curate by themselves.
What are we proposing
In light of these general considerations we are proposing that OCCSPs invest in building and maintaining a publicly accessible shared repository containing verified information on openly licensed works and works in the Public Domain.
While we believe that it is beneficial for all platforms that use copyright management tools to participate in the proposed repository, participation is supposed to have the added benefit for OCSSPs (or those platforms that could be deemed OCSSPs by the courts in the future) of ensuringcompliance with Section 18 (4) UrhDaG in particular and contributing to compliance with Article 17 (7) CDSM in general through a joint industry standard. It is necessary to define the minimum functionality requirements for the operation of the repository to fulfill OCSSPs’ legal requirements under Article 18 (4) UrhDaG, as it is the most specific provision to date regarding OCSSPs’ obligations to protect openly licensed and Public Domain works from unjustified blocking. Other Member States may copy this provision as part of their national implementations or platforms could incorporate this approach in their general compliance with Article 17 (7) CDSM.
Section 18 (4) UrhDaG states:
“After an abusive blocking request in respect of works in the public domain or works whose use by anyone is authorised free of charge, service providers must ensure, to the best of their ability and in accordance with section 1 (2), that these works are not blocked again.”
To understand this provision, it is necessary to examine (1) the nature of the obligation, (2) the conditions that trigger the obligation and (3) its consequences.
Like the obligation to block unauthorized uses of protected material, this obligation on OCSSPs to protect public domain and openly licensed content from wrongful blocking is a best effort obligation, as defined in Article 17 (4) DSM Directive and Section 1 (2) UrhDaG, respectively. That means that the obligation is subject to the principle of proportionality and must meet high industry standards of professional diligence. By entrusting a not-for-profit third party with the operation of a public repository of Public Domain and openly licensed works, to which OCSSPs add the relevant results of their complaint and redress procedures, an industry standard is established that courts will likely consider the benchmark for compliance with the best efforts obligation enshrined in Section 18 (4) UrhDaG.
For the obligation in Section 18 (4) UrhDaG to be triggered, only two conditions must be met: Firstly, there must be an abuse blocking request, that is a blocking request by a person or entity that does not hold the requisite rights to forbid the upload of a particular subject-matter. Secondly, the abusive blocking request must concern a work or other subject-matter that is either in the public domain or openly licensed. It is not relevant for the application of Section18 (4) UrhDaG whether the user-uploaded content that contains the Public Domain or openly licensed content is otherwise legal, nor whether the particular user who has uploaded the content has met all of the license requirements of an open license, such as the requirement of attribution. The OCSSP is not put in the position of having to interpret the detailed legal questions of open licensing. Rather, whenever the two conditions of Section 18 (4) UrhDaG are fulfilled (abusive blocking request with respect to a public domain or openly licensed work), the obligation to prevent future blocking of the same work is triggered. This applies regardless of whether the particular user upload that provoked the abusive blocking request ends up being reinstated, or gets blocked for different reasons.
The consequence of the obligation enshrined in Section 18 (4) UrhDaG is to ensure that the Public Domain or openly licensed works which triggered the blocking request do not get blocked again. The obligation refers to the work as a whole, not just the part of the work that the user may have uploaded. For example, if a user has incorporated a 30 second excerpt from a Creative Commons-licensed sound recording in their video and has received an abusive blocking request for the sound recording, the OCSSP must ensure that no other blocking request for any part of the same sound recording lead to the blocking of any user upload in the future – including when a different excerpt of the same sound recording is used.
This obligation confronts platforms with a practical problem: Even though users may only have uploaded a short excerpt of a Public Domain or openly licensed work, platforms are required to make best efforts to detect, and subsequently prevent the blocking of, any excerpt of the same Public Domain or openly licensed work. In order to achieve this objective, a database of the respective contents is needed that encompasses not only the exact excerpts that have led to wrongful blocking decisions in the past, but rather the entire work from which the excerpt was originally taken.
We propose the following procedure to streamline this process for all participating platforms:
An independent, not-for-profit third party is entrusted with the operation of a public repository of Public Domain and openly licensed works. The repository contains a copy of known Public Domain and openly licensed works, including relevant metadata (title, authors, copyright status, source), and identifiers such as hashes or fingerprints.
Whenever a participating platform operator learns of Public Domain or openly licensed material through a content moderation issue on its service, it sends the relevant information to the trusted third party for inclusion in the repository. The trusted third party makes the repository available to all participating platforms and the general public in machine-readable form, so as to allow the platforms to query the database as an allow-list as part of their automated content moderation procedures. By pooling the knowledge of existing Public Domain and openly licensed materials between all participating platforms, each individual platform operator will reduce the number of incidents of false blocking and the transaction costs of compliance with Article 17 (7) CDSM and specific national implementations regarding Public Domain and openly licensed content, such as Section 18 (4) UrhDaG.
One possible scenario for inclusion of new materials in the repository is that a user uploads a snippet of Public Domain or openly licensed material to an OCSSP and the OCSSP receives an abusive blocking request from a third party who is not a rightsholder in the material. Subsequent to the blocking request, the user may dispute the claim and the OCSSP will be required to subject the dispute to the complaint and redress mechanism required by Article 17(9) of the directive (implemented as Article 14 UrhDaG in Germany).
As part of the complaint and redress mechanism, the OCSSP could request a copy of or a link to the full version of the Public Domain or openly licensed work. A user who has made use of such a work will usually either be in the possession of the full work or be able to indicate a publicly accessible source. The platform then submits the full copy of the work thus obtained to the trusted third-party operator of the public repository, who may store and make the copy available to the public, as the content in question will either be in the Public Domain or published under a license that allows for inclusion in the repository.
The platform will not be liable for the accuracy of the public repository as long as it has met its best efforts obligation pursuant to Article 17 (4) CDSM. As the European Commission has pointed out in its implementation guidance, the obligation to keep legal content online pursuant to Article 17 (7) CDSM is an obligation of result and thus trumps the obligation of best effort to block unauthorized uses in cooperation with rightholders. This view is shared by the Advocate General opinion in CJEU Case C‑401/19 regarding the compatibility of Article 17 CDSM with the fundamental right to freedom of expression, pointing out that “the legislature considered that ‘false positives’, consisting of blocking legal content, were more serious than ‘false negatives’, which would mean letting some illegal content through”5. If a significant number of OCSSPs establishes a standardized procedure for establishment and quality control of the repository, courts are likely to consider a platform that participates in the use of the repository as having acted according to high industry standards of professional diligence. The accuracy of the information is expected to be high as the content in question will already have been examined by an independent person as part of the mandatory complaint and redress mechanism.
The repository could additionally be filled by platforms that are not OCSSPs, but that have comparable complaints and redress mechanisms in place as regards disputes over copyright-protected content. The repository could also be open to submissions from trusted third parties that have authoritative information about the Public Domain or openly licensed status of works, such as the Wikimedia Foundation, Creative Commons or Flickr.
The repository should be operated by an independent non-profit entity6 funded by the contributing platforms and should contain the following information that is available to the general public via publicly documented APIs as open data:
A full copy of the work (or link to a full copy of the work)
Title of the work (if available)
Relevant Standardised work Identifiers (if available)
Copyright status: “In copyright” or “Public Domain”
License (required if copyright status is “In copyright”, based on a list of open licenses)
Source and type of the rights claim (this will need a system to allow the identification of the platform that contributed the information and the method of verification. There must be the ability to record multiple sources.)
Hash(es) or fingerprint(s) (there must be the ability for platforms to include hashes/fingerprints in line with the requirements of their own ACR systems, but there should also be one default hash/fingerprint that can be used by any platform)
While anyone (including services that do not fall within the scope of Article 17) should be able to use the information published through the repository, this is not the case when it comes to adding information to the database.
For the shared repository to work it is essential that the information included in the database can be trusted by all users to be accurate. This means that the ability to add information to the repository needs to be limited to entities that can be expected to only add verified information. Since OCSSPs that would make use of the repository to comply with their obligations under Article 17 rely on the accuracy of the information in the repository, it seems logical to initially limit the ability to contribute data to platforms that make use of the repository. This could be OCSSPs that have implemented content moderation systems to comply with their obligations under Article 17(7) CDSM but also other platforms that operate content moderation systems voluntarily, including third-party content moderation service providers that are used by platforms as part of their compliance efforts. The accreditation of platforms to be able to add information to the repository could be administered by a council of contributors or a similar body.
The repository will need a conflict resolution process that can resolve conflicting information such as multiple entries that show divergent copyright status or license terms for the same underlying work. Rightsholders should also be able to contest the inclusion of a particular work in the repository, if they can show that they hold rights in the work in question that are not covered by an open license.
For this repository to become a reality it needs buy-in from a critical mass of platforms that are willing to contribute their data to the repository. This will in part depend on being able to give more precise answers regarding the technological feasibility of the proposal, assurances regarding the exclusion of liability for potentially inaccurate information, and the governance of the repository. All of these issues will need to be examined in more detail in the next step.
Regarding the technical feasibility it will be important to bring together key organisations that operate automated content recognition systems in-house (such as Google/Youtube and Facebook) and vendors that provide such technologies to platforms (such as AudibleMagic and PEX) to explore to what extent a shared repository of openly licensed works and works in the Public Domain can be included in their workflows and to define technical requirements for doing so.
Regarding the liability exclusion, it is important to involve the European Commission in deliberations to discuss the establishment of a joint industry standard of professional diligence within the meaning of Article 17 (7) CDSM.
Regarding the governance the next step would be to identify parties that could be interested in hosting the repository and/or the initial exploration phase. These could be academic research institutions, national libraries or other non-profit entities operating on the EU level.
In your presentation at the CC Global summit, you mentioned that if a user challenges a block and the rights holder does not contest it, then the content is registered in the database as PD in the presumption that if it weren’t, the alleged rights holder would have blocked it.
I wonder if this presumption of ‘PD unless successfully challenged’ could over time result in the inclusion of copyrighted material in the database, as sometimes rights holders may not contest for various reasons (too busy, too much trouble, or - indeed - if they’re not sure whether or not it is PD).
The accidental inclusion of copyrighted material in the PD DB could be counterproductive. Unless it’s irrefutably accurate, users may not feel empowered to use PD content (viz. ‘It’s probably PD, but I can’t be sure, so I won’t risk it.), and rights holders may cast doubt over it and try to discredit it. The only way to ensure that something that’s PD can be used as PD is to have a database that proves it’s PD. To repeat what you say: for this proposal to work, the PD information most be absolutely irrefutably accurate.