From Weaponising Cognitive Biases to Democratic Defence: Regulating Generative AI and Mass Media Through a Proactive EDMO

Abstract:

“Since the 2007 smartphone revolution and subsequent rise of social media, accessing information is no longer a challenge. Today, 90-95% of the global population, and 98% in Europe, access  mass media daily (ITU: Facts and Figures 2023).

However, this ubiquitous connectivity has turned the global geopolitical arena into a cognitive battlefield. Rather than exclusively fostering an informed citizenry, the modern digital ecosystem has enabled politicians and communicators of any kind to weaponise human cognitive biases. As established by Hasher and Goldstein (1977), repeated exposure to information artificially inflates its perceived referential validity.

Studies conducted by the EU highlight a similar paradox: while 79% of Europeans mention disinformation as their primary electoral concern, 61% remain highly confident in their ability to recognise fake news (Eurobarometer: Protecting Democracy, Nov 2025; Social Media Survey, Oct 2025).

The paper analyses another new central aspect of how information challenges democracy, discussing the role of Generative AI, currently widely spread in modern mass media.

Thanks to LLMs’ improvement, the use of such technology brings an increasing danger to the ability of the final user to recognise the veracity of a source presented to them.

This cognitive blind spot is actively exploited by malicious actors utilising Generative AI to manipulate political campaigns at scale (NATO StratCom COE).

Current institutional responses are insufficient. EDMO, while effective for research purposes and gathering data, remains a passive entity, lacking rapid enforcement capabilities to secure democracy and information in the mass media.

To defend democratic decision-making, this paper proposes transitioning EDMO from an observational network into a proactive regulatory authority funded by the EU, but external to it in order to prevent political leaning and corruption. By implementing mandatory algorithmic circuit breakers for unverified social media content and establishing a centralised citizen-reporting infrastructure, the EU can bridge the gap between passive fact-checking and active democratic defence.”

Introduction 

The dissemination of political information, and with it, the propagation of disinformation, is subject to constant evolution. Whereas the central challenge of electoral politics some five decades ago consisted in ensuring that voters had adequate access to information about candidates and their respective platforms, European societies today find themselves on the inverse side of that paradigm: confronting what can aptly be described as a cognitive battlefield.

Among the regulated media channels currently available, more than 95% of European citizens have access to information through a wide variety of platforms and outlets. This ostensibly reassuring figure conceals a deeper structural paradox: while 79% of Europeans identify disinformation as their principal concern with respect to electoral integrity, 61% simultaneously express high confidence in their own ability to identify false or misleading content.

The central proposition of this paper is that, as propaganda and disinformation continue to evolve in both scale and sophistication, the institutional tools designed to counteract them must evolve in parallel. In order to address the cognitive battlefield effect, the European Digital Media Observatory (EDMO) must undergo a structural transition: from a passive observational network to an independent, proactive regulatory authority equipped with genuine enforcement capabilities.

Diagnostic Foundations: Cognitive Vulnerability in the AI Era

The argument advanced in this paper is grounded in a substantial body of theoretical research on cognitive biases, the most salient of which is the Illusory Truth Effect, first identified by Hasher and Goldstein (1977) at Villanova and Temple Universities. Their research demonstrated that, when individuals assess the truth of a given statement, they rely not on critical evaluation but on whether the information accords with their prior understanding or simply feels familiar, a heuristic that can readily override rational deliberation.

A further significant finding of this research concerns the paradox of negative repetition: repeated exposure to a statement framed as false paradoxically increases its perceived truthfulness over time. The study concluded that the mere repetition of a claim substantially elevates the likelihood that it will be apprehended as factual.

Despite the longstanding nature of this theoretical framework, it retains full analytical relevance in the contemporary context. Generative AI does not, in itself, constitute the primary agent of cognitive bias weaponisation; however, it dramatically reduces the cost and complexity of producing false yet plausible information. The widespread deployment of deep-learning LLMs has rendered these systems progressively more adept at disguising falsehood as authentic content.

This cognitive vulnerability is increasingly exploited by malicious actors, particularly in the domain of large-scale political propaganda. A paradigmatic case in this regard is that of Brexit and the Cambridge Analytica affair, in which a private company successfully enabled political campaigners to deploy hyper-targeted advertising based on algorithmically derived behavioural profiles of individuals, a model that presaged the more advanced capabilities now available through generative systems.

The Paradigm Shift: Why EDMO Must Evolve

A substantiated proposal for reforming the structural framework of EDMO necessarily requires an understanding of the rationale underlying its original design. When EDMO was established in 2020, the European Union’s approach to disinformation was primarily predicated on the self-regulation of entities, whereby major privately owned mass media platforms were expected to police their own content ecosystems. This framework was adopted deliberately and proved effective in addressing the disinformation landscape of its time.

EDMO was intentionally designed as a decentralised, independent observatory rather than a regulatory body, for two principal reasons: the protection of freedom of expression and the accessibility of information, both of which constitute the foundational pillars of democratic decision-making. Prior to the widespread diffusion of Generative AI, the production of credible disinformation required considerable human effort, specialised skills, and significant temporal investment.

Generative AI has fundamentally and irreversibly altered this calculus. It has dramatically lowered the costs and technical barriers associated with producing high-quality fabrications, making the capacity to generate persuasive falsehoods accessible to virtually any actor. When EDMO’s original mandate was conceived, human-driven propagandistic disinformation still operated within the constraints of quality and velocity that a passive network of human fact-checkers could, to a reasonable extent, manage.

It is precisely this asymmetry that renders the original EDMO framework obsolete in 2026. This argument finds theoretical support in Brandolini’s Law, the bullshit asymmetry principle, which observes that the volume of effort required to fabricate a falsehood is orders of magnitude smaller than that required to refute it. Generative AI extends Brandolini’s principle to its logical extreme, dramatically compressing the cost and effort required to produce synthetic disinformation of exceptional quality.

EDMO currently operates by assessing the aftermath of political manipulation; it is structurally incapable of preventing the harm itself. By the time EDMO’s research staff publish a peer-reviewed debunking of a deepfake or synthetic narrative, millions of voters may have already been exposed to,  and potentially influenced by, that content. The EU’s Digital Services Act (DSA) itself acknowledges this paradigm, recognising that disinformation constitutes a “systemic risk” to civil discourse.

Institutional Design

The legal foundation for the proposed transition of EDMO resides in the integration of this institution within an existing European Union legal framework: the Digital Services Act (DSA). This implementation approach would render the transition process less costly, less laborious, and more legally coherent. Article 34 of the DSA legally obliges Very Large Online Platforms (VLOPs) and Very Large Online Search Engines (VLOSEs) to identify, analyse, and assess any systemic risks to the Union arising from the design or functioning of their services, including algorithmic systems, with Article 34(1)(c) specifically identifying concerns related to civil discourse and electoral processes.

Currently, the European Commission and the European Board for Digital Services (EBDS) are responsible for enforcing the DSA. However, EDMO’s current structural framework does not permit it to play any meaningful role in enforcing this existing legal architecture, for the reasons elaborated in the preceding section. To close this institutional gap, this paper proposes that the EU delegate to EDMO specific rapid-response enforcement powers, enabling it to immediately notify voters in the event of a breach of DSA provisions by political actors or politically affiliated entities operating in mass media, operationalised through the algorithmic circuit breaker mechanism described in the following section.

Repeated reference throughout this paper to the necessity of autonomous budgeting and institutional independence for a reformed EDMO reflects a fundamental concern: the risk that an EU-funded disinformation regulator might, over time, become subject to political instrumentalisation, given the presence of political actors within EU institutions. An autonomous budget is therefore a prerequisite for genuine independence.

Such budgetary autonomy should, for the same reasons, be funded entirely by the EU itself, thereby insulating EDMO from the risk of budget shortfalls being covered, directly or indirectly, by political campaign financing. The most appropriate mechanism for this would be the inclusion of EDMO in the Multiannual Financial Framework (MFF) as an autonomous agency, rather than its current status as a recipient of annual allocations through the Connecting Europe Facility (CEF) and the Digital Europe Programme (DEP). This structural change would not only provide greater financial stability and enable long-term strategic planning, but would also shield EDMO from the risk of politically motivated defunding in response to specific circuit breaker operations.

To further strengthen democratic defence standards, EDMO’s operational decisions should be subject to oversight by a multi-stakeholder board comprising members drawn from the judiciary, AI expertise, and civil rights advocacy. Sitting politicians and government representatives should be explicitly excluded from such governance structures.

This proposed institutional architecture finds precedent in two established models. The first is ENISA, the EU Agency for Cybersecurity, which was created to coordinate cross-border responses to cyber threats. The second, and more fundamental, is the European Central Bank (ECB), which represents the definitive model of European institutional independence, managing the euro and executing monetary interventions solely on the basis of economic data, insulated from political direction. A reformed EDMO modelled on these precedents would possess the credibility and structural durability necessary to fulfil its expanded mandate.

Operationalising Defence: Algorithmic Circuit Breakers

To effectively counter the weaponisation of cognitive biases without transgressing democratic boundaries, this paper proposes equipping EDMO with a centralised, citizen-driven query portal. Rather than deploying automated algorithms that would intervene directly in mass media ecosystems, an approach fraught with potential legal conflicts arising from private platform rights, this infrastructure empowers European citizens, rather than algorithmic systems, as the primary agents of verification.

Under this model, individuals who encounter suspicious or potentially counterfeit media content, particularly content relating to political information or propaganda, may submit that content directly to EDMO for rapid, independent verification. This reactive mechanism is to be legally anchored in Article 22 of the Digital Services Act, which governs the framework for ‘Trusted Flaggers’. By designating a structurally independent EDMO as the ultimate pan-European Trusted Flagger for electoral and cognitive risks within the ambit of democratic decision-making defence, civic concerns are translated into legally binding institutional action.

Upon verifying a malicious manipulation of information, EDMO would issue a formal Article 22 notice to the relevant Very Large Online Platform (VLOP) and related actors, legally obliging them to take all relevant remedial measures as stipulated by the DSA, ‘with priority and without delay’. This mechanism preserves the role of citizens as active participants in democratic defence rather than passive recipients of institutional protection.

Crucially, this citizen-centric model actively safeguards the rights enshrined in Article 11 of the EU Charter of Fundamental Rights, namely, freedom of expression and access to information, and reinforces the broader objective of democratic decision-making defence. By grounding its operations in user submissions rather than unilateral state-sponsored surveillance, the framework avoids the risks of algorithmic paternalism. Furthermore, in order to guarantee due process, any content creator subject to platform restrictions following an EDMO flag retains the right to challenge that determination through the mandatory internal complaint and dispute settlement mechanisms established under Articles 20 and 21 of the DSA.

Conclusion and Recommendations

The weaponisation of cognitive biases through Generative AI has rendered EDMO’s passive observational framework structurally obsolete. By transitioning EDMO into an independent, DSA-anchored regulatory authority, equipped with citizen-driven flagging infrastructure and rapid enforcement powers, the European Union can bridge the critical gap between reactive fact-checking and genuine democratic defence.

Secured by an institutional independence modelled on the ECB and ENISA, an empowered EDMO represents Europe’s most viable institutional safeguard against the weaponisation of cognitive biases at the expense of democratic decision-making.

Bibliography

1. Cadwalladr, C. and Graham-Harrison, E. (2018) ‘Revealed: 50 million Facebook profiles harvested for Cambridge Analytica in major data breach’, The Guardian, 17 March. Available at: https://www.theguardian.com/news/2018/mar/17/cambridge-analytica-facebook-influence-us-election (Accessed: 23 April 2026).
2. Feuerriegel, S., DiResta, R., Goldstein, J.A., Kumar, S., Lorenz-Spreen, P., Tomz, M. and Pröllochs, N. (2023) ‘Research can help to tackle AI-generated disinformation’, Nature Human Behaviour, 7(11), pp. 1818–1821.
3. Park, H.J. (2024) ‘The rise of generative artificial intelligence and the threat of fake news and disinformation online: perspectives from sexual medicine’, Investigative and Clinical Urology, 65(3), pp. 199–205.
4. Rudra, K., Ganguly, N., Mifsud Bonnici, J., Müller-Budack, E. and Manuvie, R. (2025) ‘Disinformation and misinformation in the age of generative AI’, in Proceedings of the Eighteenth ACM International Conference on Web Search and Data Mining. New York: ACM, pp. 1122–1123.
5. European Union (2022) Regulation (EU) 2022/2065 of the European Parliament and of the Council of 19 October 2022 on a single market for digital services and amending Directive 2000/31/EC (Digital Services Act). Available at: https://eur-lex.europa.eu/eli/reg/2022/2065/oj (Accessed: 17 April 2026).
6. European Council (2020) Council Regulation (EU, Euratom) 2020/2093 of 17 December 2020 laying down the multiannual financial framework for the years 2021 to 2027. Available at: https://eur-lex.europa.eu/eli/reg/2020/2093/oj (Accessed: 19 April 2026).
7. European Digital Media Observatory (2024) About EDMO. Available at: https://edmo.eu/  (Accessed: 7 April 2026).
8. European Commission (2022) Digital Europe Programme: European Digital Media Observatory. Available at: https://digital-strategy.ec.europa.eu/en/policies/edmo (Accessed: 23 April 2026).
9. European Union Agency for Cybersecurity (2024) About ENISA. Available at: https://www.enisa.europa.eu/  (Accessed: 1 May 2026).
10. Anthropic (2026) Claude Sonnet 4.6 [Large language model]. Used as an AI-assisted academic language refinement tool to improve register, syntax, and stylistic consistency of the policy brief. No substantive content — including data, arguments, citations, or conclusions — was added, removed, or modified. Available at: https://www.anthropic.com (Accessed: 5 May 2026).
11. European Central Bank (2024) Independence. Available at: https://www.ecb.europa.eu/ecb/orga/independence/html/index.en.html (Accessed: 1 May 2026).
12. European Union (2012) Charter of Fundamental Rights of the European Union, 2012/C 326/02. Available at: https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:12012P/TXT (Accessed: 2 May 2026).
13. NATO Strategic Communications Centre of Excellence (2024) Report on AI-enabled influence operations. Available at: https://stratcomcoe.org/ (Accessed: 14 May 2026).