Free Elections and “Turning Down the Noise” on Online Platforms

[Michael John-Hopkins is a Senior Lecturer in Law at Oxford Brookes University. He is the author of The Rule of Law in Crisis and Conflict Grey Zones (Routledge, 2016) and a co-author of Malign Information Psychological Operations: Concepts, Law, and Strategy (Routledge, forthcoming 2027).]

In July 2025, in Bradshaw and Others v United Kingdom, the European Court of Human Rights confronted a tension between freedom of expression and free elections. Open electoral debate depends on broad political freedom, yet that freedom may be exploited through coordinated foreign influence operations designed to undermine elections.

The applicants argued that the United Kingdom had failed adequately to investigate and protect against Russian interference. The Court found no violation, stressing the difficulty of measuring campaign effects and the absence of European consensus on the safeguards States must adopt. The judgment was nevertheless significant as it accepted that Article 3 of Protocol No 1 may impose positive obligations where hostile-State interference creates a real risk of impairing the very essence of free elections.

The UK Parliament’s Intelligence and Security Committee had already described the United Kingdom as a target in its Russia report. Importantly, it distinguished disinformation from a broader playbook, e.g. malign interference may combine manipulated narratives with covert funding, cyber operations, interference with electoral machinery, covert amplification and hack-and-leak activity. Such campaigns often seek less to persuade than to intensify division, confuse attribution, discredit reliable sources and erode institutional trust.

In Bradshaw, the evidence was sufficiently intense to be capable of impairing the essence of the electoral right, although the UK response remained within its wide margin of appreciation. Electoral protection must therefore extend beyond the ballot box to a communicative environment structured by platforms, recommenders, political advertising, synthetic media, profiling and coordinated manipulation, while keeping that environment open, plural and contestable rather than empowering the State to police political truth.

This post argues that an emerging European approach seeks to protect democratic resilience through operational denial, and potentially deterrence, rather than through official control of political truth. Human-rights law identifies the conditions necessary for free electoral choice and limits overreaction. EU and domestic regimes give practical effect to those conditions through provenance, transparency, user choice, platform risk assessment and recommender accountability. This approach also challenges the idea that transnational platforms operate in an aterritorial vacuum beyond public accountability. Responsibility does not depend solely on corporate domicile. It can instead be connected to the European users and markets served, the democratic communities in which risks materialise, and the provider’s control over the systems that generate or amplify them. These measures seek to make hostile operations harder to conceal, target, amplify and sustain. Their deterrent effect should not be assumed. Operational denial becomes deterrence only when credible, coordinated constraints reduce an operator’s expected prospects of success. 

“Turning Down the Noise” in the Digital Environment

The notion of ‘protection from noise’ must be handled carefully. It is not a right to truth, silence, comfort or freedom from disagreement. Democratic politics includes error, satire, anger and sharp contestation. The narrower concern is organised, synthetic and amplified noise, for example, engineered saturation, impersonation, microtargeting, synthetic deception, covert amplification, harassment or suppression of correction that materially impairs the ability to receive, test and contest political information autonomously. The object is campaign technique and communicative conditions, not lawful opinions or contested truths.

Judge Jakab’s concurring opinion in Bradshaw gives the idea an early Strasbourg formulation, namely excessive digital-disinformation ‘noise’ may impair the Article 10 right to receive information. The opinion brings human rights, platform governance and national security into the same analysis. In his words:

“One of the strengths of liberal democracies when compared to autocracies is that liberal democracies’ political decisions and democratic accountability mechanisms … tend to be relatively more strongly influenced by fact-based discourses weighing arguments. If we allow the digital disinformation noise levels to ravage as high as they currently are in our public debates, then we are endangering liberal democracies’ structural superiority.”

This reflects a historical reversal. The post-1989 human-rights order helped transitions towards democracy; it must now consider how established democracies should be protected when revanchist authoritarian States and resurgent anti-democratic movements exploit openness, sometimes in mutually reinforcing ways, to destabilise institutions and corrode trust. Yet protection language is itself dangerous as real, exaggerated, fabricated or unverified claims of interference may be used to delegitimise elections, discredit opponents and independent institutions, justify coercive measures and consolidate power.

Giving Citizens More Control over Online Noise

Strasbourg need not prescribe one platform model. Bradshaw confirms a wide margin in a technically complex, security-sensitive field lacking consensus. But subsidiarity does not mean that human-rights law has nothing to say. As Robert Spano has argued, process-based review gives national authorities space where they act in good faith, reason carefully and balance competing interests, while law sustains the democratic character of decision-making. Emerging practices on provenance, amplification, synthetic content, advertising and platform risk may therefore give courts firmer ground to assess informational conditions deliberately manipulated against the orientation towards “effective political democracy” in the European Convention on Human Rights.

In practice, users should know whether material comes from a real person, pseudonymous account, bot, political advertiser, influencer, campaign or verified media source. They should be able to reduce automated, recommended or paid material; choose non-personalised or following-only feeds; identify synthetic political content; and understand why they are targeted. These are questions of design and accountability, not merely content moderation. The aim is not to remove disagreement, but to make the architecture of political communication transparent, adjustable and contestable.

The Romanian Presidential Election Crisis: Protection or Regression?

The Romanian litigation shows the stakes. In Călin Georgescu v Romania, Strasbourg rejected a challenge to annulment of the 2024 presidential process because Article 3 of Protocol No 1 normally excludes presidential elections unless the office has sufficient legislative functions. It was not a merits ruling approving annulment. Its significance lies in the recorded allegations of algorithmic amplification, AI, unlabelled promotion, opaque funding and unequal campaign visibility.

Romania’s Constitutional Court Ruling No 32 of 6 December 2024 treated non-transparent promotion, misuse of social-media algorithms and AI, and undeclared campaign funding as vitiating the process. Whether necessary safeguarding or authoritarian regression, the decision marks protection from noise’s outer edge in that information-environment manipulation may be argued to affect not merely campaign fairness, but the free formation and expression of democratic choice. The Venice Commission’s urgent report supplies essential discipline as annulment should be exceptional and rest on clear competence, fair procedure, adequate evidence and a reasoned assessment that irregularities may have affected the result. Information-environment harm must be recognisable without making cancellation easy.

The cases thus establish a paired principle that sufficiently serious hostile interference may trigger positive protection, but intrusive remedies remain constrained by legality, evidence, procedure and proportionality.

Retreat from Platform Self-governance in the EU and UK

The Digital Services Act moves beyond a passive view of very large platforms and search engines. Their design and operation can generate systemic risks to fundamental rights, civic discourse, public security and electoral processes. The Commission’s electoral-process guidelines address recommenders, generative AI, political advertising, coordinated manipulation, crisis response and cooperation with electoral authorities. Common obligations and Commission supervision do not remove fragmentation as implementation still connects EU oversight with national Digital Services Coordinators and electoral authorities of uneven mandate, institutional capacity and domestic legal framework. Platforms are transnational while many public institutions responding to their effects remain territorially organised.

The Political Advertising Regulation makes visible who paid, why a voter sees a message, which audience was targeted and whether groups receive incompatible campaign versions. Transparency cannot solve manipulation, but can restore the public character of contestation. Privacy and data-protection rules, including the GDPR, constrain opaque behavioural sorting and microtargeting that place voters into tailored environments based on inferred fear, resentment, identity or susceptibility. The AI Act recognises system-design harms through rules on manipulative or deceptive techniques, vulnerability exploitation, synthetic-content transparency and AI used to influence elections. None creates a general law of malign influence; together they target opacity, behavioural steering, vulnerability targeting and lost provenance.

The UK, by contrast, relies on a dispersed legal and institutional framework rather than a single regulatory regime. It combines the Online Safety Act 2023 with data-protection and electoral law, including digital-imprint guidance and the Elections Act 2022, as well as the National Security Act 2023. The National Security Act’s foreign-interference offences cover certain coercive, deceptive or otherwise unlawful conduct affecting political processes, public functions and UK interests. The Foreign Influence Registration Scheme separately requires transparency for influence activities directed by foreign powers. These measures close important gaps, but they operate across distinct legal and institutional domains. This division of responsibility can make it harder to recognise renewed platform risks and simultaneous, cross-domain campaigns as connected activity and to coordinate an effective response. 

A limited ‘Brussels effect’ may raise global baselines as platforms, AI providers and advertisers build around transparency, provenance, audit, risk assessment and researcher access. The claim is not that EU law will become global or is beyond criticism; its influence depends on enforcement, judicial oversight, platform incentives and whether systemic-risk regulation is seen as democratic infrastructure rather than censorship. Accountability also cannot turn solely on domicile. The DSA links obligations to services offered to EU recipients, including outside providers with a substantial connection. This grounds responsibility in market access, recipients and system-generated risks while leaving fragmented enforcement and an accountability gap around platforms’ transnational power across dispersed corporate, technical and evidential structures, especially ranking, recommendation, targeting and monetisation systems through which manipulation gains visibility, velocity and scale.

Bradshaw did not create a settled right to protection from disinformation. It left States discretion because no clear consensus then existed, making it important to identify and keep under review the convergence now emerging. Across human-rights, electoral, platform, data-protection and AI law, recurring tools include provenance, transparency, recommender accountability, limits on microtargeting, synthetic-media labels, campaign-finance visibility, platform-data access, fair procedure and evidence-based remedies. Technology cannot solve deeper social, political, economic or diplomatic problems, and democracies cannot outsource electoral integrity to labels, audits or moderation tools. But neither should they leave the infrastructures shaping electoral choice without legal accountability.

Conclusion: An Emerging European Approach?

The principal development is neither a European right to truth nor one dominant regulatory regime, but a denial- and deterrence-oriented architecture. Human-rights law defines protected electoral conditions and restrains overreaction; EU and domestic law translate them into provenance, transparency, recommender design, political-advertising, synthetic-media, data-access and systemic-risk duties. These measures protect electoral functions through operational denial rather than State control of political truth as they make hostile manipulation harder to conceal, target, amplify and sustain. Operational denial becomes deterrence only if credible, coordinated constraints alter an operator’s expected likelihood of success.

In exceptional cases, this denial-oriented architecture may extend beyond transparency and platform-design measures to proportionate restrictions on the operational channels of an identified foreign influence campaign. In RT France v Council, the General Court upheld EU restrictions on specified Russian state-controlled outlets because they had a legal basis, respected the essence of expression, pursued an objective of general interest and were proportionate. The case supports bounded restrictions tied to an identified foreign campaign, not a general power to suppress ‘disinformation’. Such measures may contribute to denial, punishment or both, although their existence alone does not establish behavioural change.

Nevertheless, the approach remains fragmented as common EU rules coexist with uneven national capacity and Convention standards allowing a substantial margin of appreciation. Sanctioned outlets may evade restrictions by recreating blocked websites at new addresses, known as mirrors, or by having third parties republish their material. Foreign actors likewise obscure their involvement through apparently domestic intermediaries and shift operations to other platforms or jurisdictions when channels are restricted. This circumvention and cross-platform displacement can weaken deterrent credibility. The problem is not that campaigns are beyond territory, but that they cross territorially organised systems faster than those systems coordinate.

Protection from noise therefore means protecting the conditions in which disagreement remains visible, intelligible and contestable, not suppressing disagreement or licensing State truth control. Whether Europe’s emerging architecture deters persistent, adaptive influence campaigns will depend on closing regulatory seams without extinguishing the pluralism it exists to protect.