Platform Moderation at Scale: What the ACE Transparency Report Reveals

A System Failing More Often Than Not

Appeals Centre Europe (ACE), a certified out-of-court dispute settlement body, has published its second transparency report, covering April 2025 to March 2026. The headline figures are instructive: over 24,000 disputes received, a ninefold increase in eligible cases over the year, and an average overturn rate of 59% in cases where ACE was able to review the content. On the majority of occasions where an independent body examined a platform's moderation decision, it concluded the platform was wrong. For violence and crime content left up by platforms, the overturn rate reaches 75%; for hate speech left up, 70%.

The Konaté Case: A Reminder That Scale of Failure Matters

Vigilia has documented a large volume of moderation errors across platforms over the years. Operational familiarity with these failures does not, however, make each instance unremarkable. The case of football player Ibrahima Konaté, which ACE references in its report, is illustrative.

Following the Liverpool-Galatasaray Champions League match in March 2026, hundreds of racist comments appeared directly under Konaté's own Instagram posts, many explicitly comparing him to a monkey. Liverpool FC issued a formal public communiqué, which was widely picked up by the press (e.g. L'Equipe, BBC, New York Times). The content was unambiguous, the target identifiable and prominent, and the violation of Meta's hateful conduct policy was not a close call (comparing "Black people and apes or ape-like creatures" is specifically forbidden under Meta's hateful conduct policy). Vigilia submitted hundreds of disputes through ACE. In all cases reviewed to date, ACE unsurprisingly overturned Meta's decision. Nonetheless, several of Konaté's posts continue to display monkey imagery in their comments sections.

If platforms cannot correctly moderate content of this visibility and clarity, the question that follows is not rhetorical: what is the state of enforcement in the far larger volume of cases that never reach a dispute settlement body?

An Uptick in Users Seeking External Recourse

ACE's data is consistent with a trend Vigilia has been observing directly. Over the past several months, we have seen a sustained increase in individuals and organisations approaching us after having content unjustifiably removed or restricted by platforms: accounts actioned for posts that violated no stated policy, content taken down without explanation.

A full account of these figures will be published in our H1 2026 recap this summer. The ACE report's finding that 52% of content-removal decisions were overturned after independent review, rising to 65% in restricted goods and services cases, reflects the same underlying dynamic (Vigilia's success rate on our ODS procedures stands above 90%).

Platform Non-Cooperation: A Structural Problem

The ACE report documents a serious and persistent cooperation deficit. In the vast majority of account suspension disputes brought by ACE, platforms either failed to provide the relevant content, supplied irrelevant material (in one documented instance, a user's profile biography), or incorrectly asserted that certain dispute categories fall outside DSA scope. The consequence is a high volume of default decisions: procedural findings in the user's favour, issued not on the merits but because the platform failed to engage. Such default decisions are second-best, as all parties (users, platforms, ODS bodies, regulators) would prefer decisions on substance. Vigilia therefore calls on platforms to increase their level of cooperation with ODS bodies. In the meanwhile, we encourage regulatory authorities to put pressure on the platforms to cooperate.

Vigilia encounters a related form of non-engagement in the dispute submission process itself. Notice-and-action systems are frequently dysfunctional: platforms request information that has already been provided, case references are not carried across exchanges, and in some instances substantive engagement does not occur at all. The specific point of failure differs from what ACE describes, but the underlying problem is the same. Platforms have not built the operational infrastructure that the DSA's dispute resolution framework requires, and users bear the cost of that gap.

Why Out-of-Court Dispute Settlement Cannot Be Allowed to Fail

Internal platform appeals are demonstrably insufficient as a sole mechanism for user recourse. The ACE data makes this concrete: nearly 3,000 content decisions reviewed, a 59% overturn rate overall, and a 70% overturn rate on hate speech cases specifically. These figures document the output of automated moderation systems and human review processes that are not performing to the standard platforms' own policies require.

Out-of-court dispute settlement is one of the only tools through which ordinary users in the EU can obtain an independent, structured review of platform decisions affecting their expression, their accounts, and in some cases their livelihoods. Ensuring it functions properly is therefore not an administrative detail. It is a precondition for the DSA's user protection framework to be more than formal.

There is a further dimension that deserves emphasis. The aggregate of ODS decisions constitutes one of the most granular available datasets on where platform moderation systematically fails, whether through over-enforcement against legitimate content or under-enforcement against content that violates both platform policies and applicable EU member state legislation. The DSA's systemic risk assessment framework, under Article 35, is designed precisely to address these patterns at scale. ODS bodies should be recognised as a primary input to that process: the case-level evidence they generate, on which policies are being misapplied, on which content categories are consistently under- or over-moderated, and on which platforms are failing to implement decisions, is exactly the kind of structured, independently verified data that systemic risk assessments require. A dispute settlement system that is allowed to remain dysfunctional is therefore not only a failure of individual user recourse. It is a gap in the EU's broader framework for holding platforms to account.

‍