Vulnerabilities in information resilience during election campaigns on social media – observations from two VLOPs during the parliamentary elections campaign in Bulgaria (March–April 2026)

During the campaign period for the 2026 parliamentary elections in Bulgaria – an election with exceptionally high stakes given the ongoing political instability in the country over the past five years (with eight parliamentary elections over five years) – the most popular platforms where Bulgarian internet users have profiles and rely on as their primary source of media, informational, and political content are one of the key factors in shaping public opinion. This necessitates their intensive monitoring for practices that compromise information integrity, the conduct of fair and transparent election campaigns by election participants, and the ability of citizens and voters to make informed decisions.

This brief report outlines systemic issues observed on Facebook and TikTok – two of the most popular social media platforms in Bulgaria, with Facebook reaching approximately 65% of Bulgarian adults and TikTok reaching around 2.63 million users, or roughly 47.6% of the population – which are designated as very large online platforms (VLOPs) under Regulation (EU) 2022/2065 (the Digital Services Act, DSA), and are current signatories to the Code of Conduct on Disinformation, reporting twice yearly on their implementation commitments. On both platforms, content that violates their own policies circulates freely until it is individually reported to the platform with explicit reference to those policies. 
 

Issue 1: Political advertising and the trading of Facebook groups

An established practice of buying, selling, and repurposing Facebook groups is actively exploited during election campaigns. An established practice of buying, selling, and repurposing Facebook groups is actively exploited during election campaigns. The sale and transfer of Facebook groups explicitly violate Meta's own Community Standards on Spam, which prohibit "attempting to or successfully selling, buying or exchanging platform assets, such as accounts, groups and Pages".

Regardless of the formal potential for groups to transition from one topic to another, this practice creates conditions for heightened commercial interest in group audiences and opportunities to direct users who are active in a given group towards the positions of a particular political party – without their knowledge or consent, and without them having had the opportunity to assess and explicitly agree to such a change. A DFRLab investigation published in March 2026 documented a coordinated Bulgarian Facebook network – comprising two pages, nine groups, and six accounts, with the nine groups alone totalling over 211,600 members – systematically driving traffic to a fabricated, monetized news site. Factcheck.bg had documented a related ecosystem of over 100 Facebook groups and pages involved in spreading disinformation since 2020. Monitoring conducted by the Balkan Free Media Initiative (BFMI) during the 2026 Bulgarian elections also identified boosting through purchased pages and coordinated inauthentic cross-platform activity as active tactics.

Under Article 34(1) of the DSA, VLOPs are required to "diligently identify, analyse and assess any systemic risks in the Union stemming from the design or functioning of their service," including algorithmic amplification and platform features that may enable or exacerbate information manipulation. Article 35(1) DSA requires that VLOPs "put in place reasonable, proportionate and effective mitigation measures" tailored to the systemic risks identified, with specific attention to impacts on democratic processes and electoral integrity. The European Commission's guidelines for VLOPs on the mitigation of systemic risks for electoral processes, published in April 2024, identify Articles 34–35 as the primary provisions governing platform conduct during electoral periods. The coordinated repurposing of audience-bearing groups for political purposes – a practice with documented recurrence across multiple Bulgarian election cycles – falls squarely within the category of risks that these provisions address. The question of whether VLOP risk assessments have adequately captured this specific mechanism, and what mitigation measures have been put in place in response, is one that supervisory authorities and civil society researchers are in a position to ask.

Issue 2: Meta's suspension of political and social issue advertising

On 24 July 2025, Meta announced it would cease to allow political, electoral, and social issue advertising on its platforms in the EU from early October 2025, citing "significant operational challenges and legal uncertainties" introduced by Regulation (EU) 2024/900 on the transparency and targeting of political advertising (TTPA), which it described as containing "unworkable requirements". Meta also stated that the TTPA "will restrict how advertisers can reach their audiences," signalling concern about the regulation's impact on its personalized advertising model alongside its compliance concerns. The TTPA entered full application on 10 October 2025.

The TTPA's core obligations include mandatory labelling of political advertising, disclosure of the sponsor's identity and financing, restrictions on targeting based on sensitive data categories, and the maintenance of an advertising repository accessible to the public and to researchers. By opting out of the political advertising market rather than implementing these obligations – without seeking binding regulatory guidance or challenging the regulation through legal channels – social media companies like Meta traded a transparency regime for opacity. The formal, attributable, regulated channel for political communication on Facebook was closed, while informal and organic channels – including the group networks described above – continued to operate without adequate transparency infrastructure. This outcome is in direct tension with the TTPA's stated purpose of protecting the integrity of elections and democratic participation.

It should be noted that the regulation's definition of "political advertising" is acknowledged to be broad and, in some respects, difficult to apply consistently at scale, and independent legal analysis has identified genuine ambiguities in the identification regime, particularly where the primary obligation to declare an advertisement as political rests with the sponsor rather than the platform. These substantive regulatory design issues, however, do not diminish the transparency deficit created by the suspension – they underscore the need for clearer regulatory guidance and enforcement, not the removal of the advertising category.

Issue 3: AI-generated content on TikTok not labeled

Another trend is accounts identified that are posting deceptive and derogatory AI-generated content, the majority of which concerning politicians – without the proper labeling. Given the high volume of such content that users are confronted with on a daily basis across platforms, and that it is becoming increasingly difficult to recognize such content by the average user, the labelling needs to be more systematic, as piecemeal detection and actioning is not effective. Users need to be better informed about AI-generated content. Detecting such content by independent monitoring actors is impossible at scale. 

Issue 4: Sanctions circumvention: content from sanctioned Russian media outlets  

Content sourced from Russian media outlets subject to sanctions under Council Regulation (EU) No 833/2014 was found on both TikTok and Facebook, which often is mixed up alongside local political messaging posts, which makes it highly relevant in the election context, but also deserves attention on its own as illegal content and how systematically it is actioned. While such content flagged was actioned via geo-blocking by TikTok (restricting the visibility of this content in the country flagging it), it remains visible in other EU Member States, and is still reaching voters living abroad (when assessing the relevance in the election context), while wider EU audiences still have access to sanctioned content. There is currently no mechanism to systematically action such content across all Member States, beyond piecemeal national-level flagging. 

Similar content was also found on Facebook – either directly posting entire news segments from various Russian TV channels subject to sanctions (identical to the case flagged to TikTok), or as previously described in this report – content from EU-sanctioned Russian entities being translated verbatim and republished at scale by domestic media outlets such as Pogled.Info – which is then shared on Facebook. 

The issue of sanctions circumvention is not confined to a single platform or Member State and mandates a more effective and holistic solution.

~

Obligations under Article 28b AVMSD and media literacy

Directive (EU) 2018/1808 (amending Directive 2010/13/EU, the Audiovisual Media Services Directive, AVMSD) introduced specific obligations for video-sharing platform providers (VSPs) through Article 28b. Both Facebook and TikTok meet the criteria for VSP classification under the AVMSD on the grounds of essential functionality. A critical structural feature of Article 28b is that it imposes obligations on member states to create a legal environment in which VSPs under their jurisdiction take appropriate measures – it does not impose direct obligations on VSPs as such, since the AVMSD is a directive rather than a regulation, and VSP compliance arises from national transposition. Primary supervisory responsibility for both Meta and TikTok falls under Irish jurisdiction (Coimisiún na Meán), not a Bulgarian authority. This does not diminish the relevance of Article 28b to the issues described in this report, but it does affect the question of enforcement standing and escalation channels.

Among the measures enumerated in Article 28b(3), two are of particular relevance here. Article 28b(3)(d) requires VSPs to establish and operate "transparent and user-friendly mechanisms for users of a video-sharing platform to report or flag" content that violates the platform's obligations. Article 28b(3)(e) requires platforms to "establish and operate systems through which video-sharing platform providers explain to users what effect has been given to the reporting and flagging" under point (d). Article 28b(3)(j) explicitly requires VSPs to provide "effective media literacy measures and tools and raise users' awareness of those measures and tools".

The issues documented in this report expose a structural gap in how these obligations are currently met in practice. Content that violates platform policies continues to circulate until individually reported by users, placing the burden of enforcement on individual citizens. This presupposes a level of familiarity with platform policies, reporting mechanisms, and the DSA's complaint infrastructure that most users – and particularly users in media literacy contexts that remain under-resourced, as in Bulgaria – do not possess.

The obligation under Article 28b(3)(j) to provide effective media literacy tools, read alongside the transparency and reporting obligations in points (d) and (e), points toward a standard that goes beyond making reporting buttons technically available: it requires that users be genuinely informed about what constitutes prohibited content, how to exercise their right to report it, and what to expect as a result.

The current design of platform policies – fragmented across multiple help sections, written in generic legal language, and rarely surfaced in the contexts where users encounter violating content – falls well short of this standard. In the specific context of election campaigns, where the pace and volume of content production make individual reporting an inadequate safeguard, the argument for proactive platform-level action, supported by robust user-facing information about platform rules and DSA-based remedies, is well-grounded.

~

Bulgaria's activation of the DSA Rapid Response System – a voluntary mechanism that connects platforms, civil society organizations, and fact-checkers, operating within the DSA framework and the Code of Conduct on Disinformation, confirmed by the European Commission on 24 March 2026 for the period up to one week after the 19 April elections – follows similar activations in Romania and Portugal. 

The Rapid Response System, however, cannot be and is not intended to be a sufficient safeguard on its own: a review of platform risk assessments for recent European elections found that platforms consistently asserted the adequacy of their mitigation measures, while external civil society observations documented persistent failures to detect political advertising or identify election-related manipulation; and, critically, the DSA and the Commission's election guidelines do not specify operational benchmarks for what "reasonable, proportionate and effective" mitigation measures must look like in an electoral context. The measures outlined above – transparency in advertising markets, platform-level action on the group economy, and genuinely accessible user information and media literacy tools as required by Article 28b AVMSD – constitute the substantive layer that a notification mechanism alone cannot replace.

~

See also: EDMO’s assessment of the Rapid Response System of the Code of Practice on Disinformation