Eighteen months after Romania's Constitutional Court took the extraordinary step of annulling the first round of its 2024 presidential election, the European Commission's formal proceedings against TikTok under the Digital Services Act remain open — and remain, in our view, one of the most consequential regulatory experiments anywhere in the democratic world. Brussels is not merely auditing a recommender system. It is testing whether a hub-and-spoke rulebook for the open internet can address election-integrity concerns without sliding into the kind of state-directed speech control that the DSA was explicitly designed not to be.
The case began on December 17, 2024, when the Commission opened proceedings against TikTok under Articles 34, 35 and 38 of the DSA, citing concerns about systemic risks tied to electoral processes and the platform's recommender systems and political advertising policies. The trigger was unmistakable: only days earlier, Romania's Constitutional Court annulled the first round of its presidential election after the country's Supreme Council of National Defence released declassified intelligence alleging coordinated inauthentic behaviour favouring far-right candidate Călin Georgescu. As of mid-2026, the probe is still active, with the Commission continuing to demand technical disclosures from TikTok and exploring whether the platform met its risk-mitigation obligations during the campaign.
What the DSA Actually Requires
It is worth restating what the DSA does and does not do, because public commentary has often blurred the two. Under Article 34, Very Large Online Platforms (VLOPs) — those with 45 million or more monthly EU users, a designation TikTok received in April 2023 — must conduct annual risk assessments covering, among other things, “actual or foreseeable negative effects on civic discourse and electoral processes.” Article 35 obliges them to deploy “reasonable, proportionate and effective” mitigation measures. Article 38 imposes transparency around recommender systems, including offering at least one option not based on profiling.
Crucially, the DSA does not authorise the Commission to order the removal of lawful political speech, nor to dictate which candidates a platform should amplify or suppress. The legal hook is procedural: did TikTok identify the risk, did it design adequate mitigations, and did it document them? That is a defensible, evidence-based standard. It is also a standard that requires regulatory restraint when applied to contested political content.
The Romanian Backdrop
Romania's annulment was unprecedented in modern EU history. Whatever one thinks of the Constitutional Court's reasoning — and many serious commentators, including civil liberties groups, have questioned the proportionality of voiding an election on the basis of platform-amplification claims — the underlying intelligence pointed to networked behaviour rather than the substance of any single user's speech. That distinction matters. Coordinated inauthentic behaviour, paid amplification dressed up as organic content, and undisclosed political advertising are the legitimate targets of platform regulation. A candidate's opinions, however unwelcome, are not.
TikTok, for its part, has reported removing networks of accounts during and after the campaign and published transparency disclosures on its electoral integrity work. Whether those measures were timely and proportionate to the documented risk is exactly the kind of question the DSA process is designed to answer.
The Censorship Trap
The risk in this case is not that the Commission will be too tough on TikTok. It is that the proceedings will set a precedent in which “systemic risk” becomes a flexible category that regulators reach for whenever an election produces an inconvenient result. Civil society groups including the Electronic Frontier Foundation have warned across multiple EU digital files — most recently in EFF's May 2026 submission on the Digital Fairness Act — that broad, outcome-oriented mandates for online services tend to drift toward over-removal once enforcement begins. The same dynamic threatens the DSA.
If the Commission's final findings hinge on the content of pro-Georgescu videos rather than on whether TikTok's processes were adequate, the message to every VLOP will be: when in doubt, throttle. That outcome would harm smaller political voices far more than the established ones, narrow the space for legitimate political insurgency online, and hand authoritarian governments outside Europe a ready-made template for compelling platforms to suppress disfavoured speech in the name of “election integrity.”
A Proportionate Path Forward
A pro-innovation, pro-speech enforcement of the DSA in this case would focus on four concrete questions, all squarely within the regulation's text:
- Disclosure of paid political content: Did TikTok adequately enforce its own ban on paid political advertising, including against creators receiving undisclosed compensation?
- Recommender system transparency: Can TikTok demonstrate that its “For You” ranking signals during the Romanian campaign were not gameable by coordinated networks in ways the company knew about and failed to mitigate?
- Researcher access: Under Article 40, has TikTok provided vetted researchers with the data needed to reproduce the Commission's findings independently?
- Process, not viewpoint: Are remedies tailored to authentication, labelling, and provenance — rather than to amplification penalties for content based on its political valence?
The DSA's promise was always that it would replace ad-hoc takedown pressure with predictable, auditable obligations. The TikTok investigation is the first major test of that promise in an election context. Brussels should resist the temptation, however politically attractive, to convert a process-based rulebook into a content-based one. The credibility of the entire DSA architecture — and Europe's claim to a rights-respecting alternative to the surveillance-state model of internet governance — depends on getting this distinction right.