Platform accountability in the EU has a language blind spot

LSE Media Policy Project · 2026-09-03 · UK · verified record

Themes: Child safety & minors · Online platforms · Search services

Monika Wallmon of the University of Gävle, Sweden explains how EU attempts to make platforms accountable for online content moderation do not fully take into account all languages spoken in the bloc, such as the Sámi languages. The European Union’s Digital Services Act (DSA) has applied across the EU since February 2024. It is an important regulatory advance; it strengthens transparency, requires the largest platforms to assess systemic risks and gives users ways to challenge platform decisions. In February 2026, something genuinely new followed: the first transparency reports on online content moderation using a common EU-mandated format. The harmonised, machine-readable template has been in use since July 2025. It makes reports comparable across platforms and reporting periods. For very large online platforms, it requires figures on moderators with sufficient expertise in each official language of an EU Member State and on the accuracy, precision and recall of automated moderation in those languages. Very large online platforms must therefore disclose their reported linguistic moderation capacity in a form that others can scrutinise. But the same reporting architecture that makes moderation more comparable can leave linguistic inequality unmeasurable within the reports themselves. It reports language-specific capacity and automated-system performance as inputs. It also counts internal complaints, whether platform decisions are upheld or reversed and how long decisions take, but not by language. Regulators can therefore see what resources a platform declares, but not whether the rights and remedies created by the DSA are equally usable across languages. In these reporting tables, a language becomes visible in the data only if it appears as a separate category. Inclusion is not a service promise, and omission does not prove that support is absent. Yet the template’s categories are organised around the official languages of EU Member States. It can compare declared capacity in Swedish and Finnish, for example, but provides no equivalent category for Sámi languages or other languages outside this framework. Complaint outcomes, appeal outcomes and waiting times, meanwhile, are not broken down by language at all.…