Art. 21 DSA - What to expect?

Art. 21 sticks out from the rest of the provisions of the new EU Digital Services Act (DSA). While much of the DSA either copies existing (national) regulation (transparency, complaints mechanisms, data access), or seems more like symbolic legislation (trusted-flagger, repeat-infringer rule), or is ambitious but also painfully vague (the heart of the DSA: Its risk mitigation regime), Art. 21 is an extravagant provision: it is an unprecedented approach, it is very detailed and complex and, nevertheless, it will spur the establishment of an entirely new settlement industry. 

For some time now, I have discussed with various stakeholders what to expect with this new provision. Here are my key take-aways from this plus some thoughts what I think to expect:

Before I turn to rumors from stakeholders and predictions, let me dive into a short explainer on what the provision does and not:

Art. 21 allows private actors (lawyers, for example, would be a good fit) to seek certification as out-of-court dispute settlement bodies for all kinds of content moderation decisions under the DSA. That is, e.g., if an online platform does not take down content notified as hate speech, or, from an Uploader’s perspective, if a platform suspends an account for allegedly violating its Community Standards, one can then turn to such a certified body and seek review. The Art. 21 body then will reach out to the user as well as to the platform. It might then try to reach a settlement. However, if it does not, it is expected to make a court-like decision about the legitimacy of the moderation decision in question.

Platforms will have to engage, in good faith, with the Art. 21 bodies (that is: Users can drag the platform into the forum!). But according to the wording of the new law, the decisions then taken by the bodies are (except for the costs, see below) not binding (the draft version of the DSA wanted to enable binding decisions, see critique here).

However, scholars already argue that there will still be some kind of binding effect, that “unjustified ignorance on substance can trigger consequences” (Husovec, Martin and Roche Laguna, Irene, Digital Services Act: A Short Primer). I doubt that substantial consequences can follow (see my treatise in Müller-Terpitz / Köhler, Digital Services Act: DSA (Oct. 2023), Art. 21). But this is obviously a hot topic (a question that we can ask the Commission or the DSCs in the future).

However, Art. 21 will attract a lot of applicants (users seeking an external review over a platform moderation decision). And the reason for this is the asymmetrical cost scheme that Art. 21 enables: In most of the cases, even if the platform “wins” (that is: the settlement body confirms the moderation decision), the platform - at least according to the most convincing interpretation of the law - still will bear all or most of the costs of the procedure. In my view, this cost-scheme is a core legislative mistake in Art. 21 (see my critique here and in-depths in Müller-Terpitz / Köhler, Digital Services Act: DSA (Oct. 2023), Art. 21). 

Anyway, this cost-scheme is exceptional: Just imagine a world where when you have a dispute with a contracting party, e.g. your landlord raises your utility costs, you can drag the landlord before a settlement body which delivers a non-binding “decision”, at the costs of the landlord - even if the landlord “wins”. Now keep in mind: Art. 21 will apply to millions of content decisions, in an area where we already are seeing platform-users filing millions of in-house appeals (“My post was not hate-speech, restore it!”). And initiating the Art. 21 proceedings will be just a mouse-click away. I expect that large numbers of users will soon start to flood the Art. 21 - settlement bodies, causing the platforms a lot of costs, and - this is my prediction - incentivizing a flourishing settlement-body-industry (if you are a young lawyer looking for a business model - why not get certified?). Especially users who like to defend their deleted “hate speech” or conspiracy theories will predictably be happy customers of the new bodies.

How can you become a certified settlement body? You have to apply with your national Digital Services Coordinator (DSC), Art. 21(3) DSA. However, many Member States - e.g. Germany - have not yet appointed a DSC, as the final deadline for this is February 17th 2024, Art. 49(3) DSA. But theoretically, Art. 21 already does apply today, for very large online platforms (VLOPs), see Art. 92 DSA. So what do you do, if there is the law, but no DSC yet? You might turn to the European Commission, who, according to Art. 56(3) can enforce the whole DSA for VLOPs (very large online platforms), also in areas outside the risk mitigation scheme  like Art. 21 where otherwise the national DSCs would be in charge (Art. 56(4)). Already the Commission seems to take up such a role for other provisions of the DSA - at least in parts for Germany - as an interim measure (see here). However, I know of no stakeholders ready for immediate certification, so we will probably see the first applications only in 2024.

Potential settlement bodies might think twice about choosing their Member State of establishment, e.g. to opt for a country with a “certification-friendly” DSC (e.g. as a measure to attract such institutions) - as once you are certified, users from all over Europe can bring cases (forum-shopping, Art. 21(1)).

One last look into the crystal ball: How long might certification procedures take? This is hard to answer. One might expect great differences amongst Member States here. In bureaucracy-loving Germany, the similar task to certify FSM as a self-regulator under the NetzDG took pretty long (application end of 2018, final start as self-regulator in 2020).

When you deal with Art. 21, the parallels to the (Meta) Oversight Board are obvious. This might cause some strange competition for the well-funded (and pretty exclusive) Oversight Board. However, since spring 2023 rumors grew louder and louder that the Oversight Board itself might in some form (probably through a separate body or in an affiliated setup) seek certification. This could be very much in the interest of Facebook and Instagram, because these platforms then would find a body they trust and where they could communicate willingness to accept decisions.

The FSM, which is a certified self-regulatory body under the NetzDG, might also be seen as a role-model for Art. 21 DSA. And - as expected - its work has come to a halt under the DSA (as NetzDG will vanish). However, more than 60 lawyers had worked for the FSM as (part-time) self-regulators under the NetzDG. They obviously have suitable expertise to now seek certification under Art. 21 DSA. But people familiar with planning there have said that while the idea of the FSM seeking certification as an organization is certainly interesting, but  decision making would not be so easy. A core issue is how to plan financing, as any substantive setup would result in basic costs just for setting up the organization - while it might be difficult to predict how many cases will come to refinance these costs through the procedural fees that come with every single settlement procedure.

In my view, lawyers (especially in the field of personality rights, intellectual property, media law) are a natural fit to run entities qualifying for certification under Art. 21. And yes, there has been feedback to me that lawyers are willing to seek certification. E.g. rumors are that some lawyers who have worked in the FSM self-regulation are considering whether to seek certification on their own. Beyond that, my impression is that Art. 21 DSA is not yet very much on the radar for lawyers as a potential business model.

According to Art. 21(6) DSA, Member States themselves may establish out-of-court dispute settlement bodies or support them (e.g. financially). My prediction had always been that this would not happen, as it seems hard to justify spending taxpayer money on this, in the end these disputes are of private nature, and users already have easy access to courts, which regularly and skillfully review content moderation decisions. That is why I found it very surprising that in early 2023 a (leaked) draft version of a German law for implementing the DSA included the establishment of a government-based dispute-settlement body (see here). However, this idea of a state-run body is not included in the official first draft version of the law any longer. 

Some stakeholders that I talked to were thinking about how to set up solutions at scale through legal tech, e.g. through automated “settlement”-decision making (or support, e.g. through AI-analyzing of content-moderation decisions etc.; the machine-readable database to be set-up under Art. 24(5) DSA might be helpful). Business ideas vary from running such entities as legal tech start ups or providing technical solutions as a background provider for certified settlement-bodies.

One of the biggest riddles of the DSA is its scope and relationship with existing regulation. As with most modern EU-legislation: The interplay with existing regulation becomes so complex, that legislators favor leaving these questions effectively unsolved (may the courts make sense of it!).

In the area of how rights holders or users can appeal content moderation decisions they are unsatisfied with, the DSA will have interesting interplay with AVMSD, DSM-Directive, P2B-Regulation etc. This is a complex topic, I am only scratching the surface here. However, I want to highlight that most scholars argue that for Art. 21, the DSA will be more specific and override Art. 17(9) DSM-Directive. So in copyright matters, e.g. a rightsholder demanding a take-down, or a user insisting on a rightful publication and requesting put-back, Art. 21 DSA will introduce a new mechanism: You probably can bring these disputes now to the settlement-bodies. This many result in weird outcomes: deep-pocket film studios or other commercial actors will now have another button to push - escalating disputes to the Art. 21 - settlement-bodies, at a platform’s cost.

I have been talking to platform lawyers and lobbyists since the draft versions of Art. 21 DSA. During the legislative process, they saw the headaches coming. However, while there had been interventions, it had not been a number one priority (the attitude was more like: Well, it's gonna be costly and burdensome, but we have to focus on other topics of the DSA). Now that the DSA has been passed, it seems that in-house lawyers are still struggling to define a strategy on how to deal with the future Art. 21 proceedings. My impression was that platforms expect that they cannot plan with just following decisions, however, some platforms might be interested in identifying certain trustworthy actors (e.g. this could be the Oversight Board or FSM) where they could implement a general strategy of opening up to “more likely” following the decisions or where they could establish trust in reasonable outcomes.

Another question that I personally find interesting is whether platforms - which in the end, have to finance the settlement bodies - will be in a formal position to seek review of dubious certifications before administrative courts. In my opinion, this should be the case. Another - more theoretical - question is whether platforms could attack the cost-scheme of Art. 21 itself. Platforms could argue that forcing them to potentially finance millions of non-binding external decisions is disproportionately affecting their freedom to conduct a business, because there is already an escalation path for external review available: Judicial review in the courts. I think there is a case, but I admit to be pretty alone with this heavyweight criticism of Art. 21 (see my critique in Müller-Terpitz / Köhler, Digital Services Act: DSA (Oct. 2023), Art. 21).

One last thought: Art. 21 could worsen platform content moderation. Here is why: Art. 21 puts a “price tag” on moderation decisions (which might have been intended by the legislators, as is indicated by respective academic arguments which might have played a role in drafting Art. 21). The problem: Through its asymmetrical cost scheme (see above 1.3.) the Art. 21 - “price tag” kicks in not only for “wrong” moderation decisions, but - in general - also when the platform got it right (remember: The platform pays for the procedural fees even if it “wins”). This might disincentivize platforms to take-down illegal content when they predict a likelihood of users turning to settlement bodies (my forecast: Haters and conspiracy theorists will love to initiate Art. 21 proceedings). Such financial disincentives to take moderation decisions then might lead to underblocking in areas where proper enforcement of the Community Standards should be welcomed. In similar situations, it might make platforms less willing to defend perfectly fine moderation decisions (to prevent the costs of Art. 21) - a situation one could describe as over-put-back. So Art. 21 might not lead to better content moderation, but worsen it (see my critique here and in-depths in Müller-Terpitz / Köhler, Digital Services Act: DSA (Oct. 2023), Art. 21). For the moment, such chilling effects of Art. 21 are theoretical - but should carefully be evaluated.

Now, after all: Who oversees the settlement bodies, or - in a broader sense - has the authority to shape the way these private entities will work? Who will ensure that bad actors cannot enter or will have to leave the market?

Theoretically, the most effective tool to make sure that only “good” settlement bodies enter the market is the certification process. DSCs shall only certify a body if it meets the conditions laid down in Art. 21(3). Most relevant, in my opinion, are the required conditions of impartiality, independence and expertise. This seems crucial, as in my view, Art. 21 - bodies will face an in-built bias to make decisions that in doubt favor users (reason: To be attractive for users who are the only ones initiating the process, thus securing financing through procedural fees). The certification process probably can try to focus on safeguards in this regard, but only to a certain extent. 

Another question therefore is how to prevent a situation where user-friendly settlement bodies attract thousands of irrelevant cases just to keep their own financing running, which might disproportionately burden platforms with the cost of irrelevant proceedings. In my opinion, a true gamechanger in that regard would be DSCs requiring that the bodies generally request a substantive minimum nominal fee from the applicants (see Art. 21(5)). This would prevent thoughtlessly initiated proceedings. 

However, I am afraid it is over-optimistic to rely on the DSCs applying a strict enough certification regime. The main problem: Every DSC of every Member State can certify bodies which will be eligible to work in the whole Union. Even if just a few DSCs do not follow a good enough approach the damage is done. In my view, the Commission would be wise to push for strict certification requirements. More formally, the European Board for Digital Services, which has the role to support a consistent application of the DSA, could contribute greatly by focusing on this topic.

Another important role for national regulators is the ex post oversight over settlement bodies: Settlement bodies have to report to the DSCs annually, and a DSC will revoke certification if a body no longer meets the criteria which had been relevant for certification, e.g. impartiality, independence and expertise. Now you might think: Voila - If a settlement body delivers bad judgments (or has a habit of one-sided rulings) then it will be decertified. But I am very skeptical that this scenario will lead to substantive outcomes or incentives. DSCs would have to review a large number of decisions, and then would need to make a case that the decisions were “too bad”. But this will be burdensome and difficult, also because the settlement bodies are expected to be “independent”. What an immense task! Even if a DSC could show that a certain body did deliver a lot of “not so convincing” interpretations of the law - this will not prove the body lost its expertise! And the settlement bodies will fight de-certifications in administrative courts! Plus: Revoking the certification of a body that had a reputation for deciding cases too “free speech friendly” might also be a politically hot topic. So to sum it up: I personally do not have much faith that we will see substantial ex post oversight over the quality of Art. 21 bodies.

Another form of oversight might come from courts. As I have already mentioned, I think that platforms should be in a position to seek judicial review of certification (see above Nr. 9). However, the bodies’ decisions on specific cases will likely not find review in courts, as they are non-binding on the matter, thus parties would likely have no standing (what do you sue against?). One exemption: Costs! Platforms could reject overpriced “fees” charged to them or (un)reasonable expenses that users ask to be reimbursed under Art. 21(5).

Art. 21 is potentially disruptive. Nevertheless, Art. 21 is still relatively under the radar of many stakeholders. This will change as large numbers of cases might be incoming, nourishing lucrative business models. As a matter of concern, the construction of Art. 21 does not provide for sufficient safeguards against  “bad actors” taking advantage. However, from 2024 on, I expect more and more entities seeking certification, and legal tech start-ups to offer scalable solutions (AI-supported settlement?). It is still hard to guess how platforms will respond to a flood of proceedings they then have to engage in. One thing is for sure, Art. 21, and what stakeholders do with this new approach, requires our attention. Art. 21 is - for the better or worse - not like many other new regulatory approaches.