A state monopoly on truth

Analysis and commentary based on material supplied to the editorial team. Sources checked on 5 November 2026. “A state monopoly on truth” describes a risk and a critical concept, not an established legal monopoly in the European Union. Conceptual illustrations were generated using AI.

Disinformation is a real threat. That does not give governments or platforms unlimited authority to decide which voices deserve a place in public debate. The central dispute now concerns the visibility of information as well as its removal. Can regulation constrain technology companies without creating new instruments of pressure?

Introduction: who decides what we see?

The internet promised to weaken information monopolies. Anyone would be able to publish without permission from a publisher, broadcaster or government office. Publishing did become widely accessible. Access to an audience remained uneven: search engines, recommendation systems and private platforms’ rules now stand between speakers and readers.

This shift has political consequences. A post may remain accessible through its direct address while disappearing from recommendations or search results. Its author can still speak, yet reach fewer people. The distinction between removal and restricted distribution can therefore matter greatly in law while remaining almost invisible to an audience.

Not every such change is censorship. Information has to be ordered, and no recommendation system can give every statement equal reach. The problem arises when restrictions are opaque, arbitrary, discriminatory or imposed under government pressure without an appropriate legal basis and oversight. Situations in which public regulation, a platform’s commercial interests and external verification overlap are particularly difficult to assess.

The argument of this article is therefore conditional: protecting people from disinformation can become an instrument of improper control over public debate when transparency, proportionality and effective remedies are missing. The existence of regulation alone does not prove this allegation. It requires examination of particular decisions, their grounds and their consequences.

Part I. A real problem that requires careful measurement

1.1. Disinformation, mistakes and opinions are different categories

Debates over regulation often merge deliberate deception, the unintentional sharing of inaccurate information, satire, interpretative disagreement and illegal content. This makes it harder to judge proposed responses.

For this analysis, disinformation means deliberately misleading an audience. Misinformation refers to false or misleading information shared without necessarily involving that intention. An opinion, even a harsh or unfair one, is not automatically false information. Meanwhile, threats, fraud and unlawful disclosure of personal data can violate the law whether or not the statements involved are true.

These distinctions should precede intervention. Removing financial fraud, labelling a politician’s inaccurate claim and reducing recommendations for lawful material require different justifications. The word “harmful” cannot substitute for identifying the harm and establishing who has authority to assess it.

1.2. A risk ranking does not measure the extent of damage

In the World Economic Forum’s Global Risks Report 2026, misinformation and disinformation rank second over the two-year outlook to 2028. This reflects experts’ perceptions of risks, rather than a statistical measurement of false publications or their electoral effects. It should not be confused with the report’s separate ranking for crises in 2026 itself. [1]

The ranking indicates concern but does not establish which legal instruments are necessary. A serious problem does not automatically justify every block or moderation requirement. Evidence of effectiveness and an assessment of interference with rights must connect the diagnosis with the response.

1.3. $417 billion: an important estimate, not an exact loss account

In March 2026, Sopra Steria published an assessment estimating the global economic impact of disinformation in 2024 at approximately $417 billion. It combines previous research, documented cases and economic modelling. Its presentation gives scenarios ranging from $356 billion to $516 billion. [2]

The meaning of individual categories is crucial. The $227 billion associated with fake reviews refers, according to the authors, to consumer spending influenced by those reviews. It is not equivalent to a demonstrated loss of that amount. Nor should the entire $417 billion be treated as an exact recurring annual reduction in global GDP. [2]

Such models can help convey scale, but combine phenomena with different levels of measurability. Fraud, advertising revenue and loss of trust are not identical economic categories. Adding them requires scrutiny of assumptions and potential overlaps. The publisher’s perspective also matters: a technology company describing the market for information resilience is a market participant, rather than a statistical authority.

A critical examination of disinformation should apply the same standards to its own figures that it demands of others. An impressive number without methodological explanation can hinder debate as much as an absence of data.

1.4. AI expands manipulation, but fact-checking data have limits

According to EDMO, AI-generated material accounted for 20% of detected disinformation in March 2026 in the monitoring described by the observatory. EFE Verifica reported that 34% of its checks that month concerned synthetic content. These figures describe material detected and selected for checking, rather than a representative sample of the internet. [3] [4]

Geography matters too. BOOM, cited in discussions of rising AI content, operates in India, not Spain. Its review of 2025 reported AI-generated content in 20.5% of checked material, compared with 8.35% in 2024. The comparison concerns its own verification work. [5]

Synthetic content does not exhaust the problem. An authentic video supplied with a false date or caption can mislead without any generated imagery. Conversely, the ability to produce deepfakes makes it easier to challenge genuine evidence: one can simply suggest that an inconvenient recording was generated. EFE also describes this mechanism. [4]

The conclusion is not that AI fabrication requires a single arbiter of authenticity. We need transparent documentation of provenance, access to originals and willingness to correct mistakes. A label should help readers check information, rather than end a discussion through institutional authority alone.

Part II. What does the regulatory architecture actually require?

2.1. The DSA does not impose a general prohibition on untruth

The Digital Services Act is an EU regulation, rather than a directive whose entire substance must be transposed into national law. It has generally applied since 17 February 2024. Member states nevertheless have to provide institutions and procedures for enforcement. A delayed national statute does not mean that the regulation itself did not apply before it. [6]

The DSA does not make every inaccurate statement illegal. It distinguishes obligations concerning unlawful content from management of broader risks by the largest services. It also prohibits imposing a general obligation to monitor everything users publish. The maximum fine of up to 6% of global annual turnover concerns breaches of obligations under the regulation; it is not an automatic penalty for leaving a single controversial opinion online. [6]

It is easy to overlook the regulation’s two directions. It can require a response to illegal material while also constraining arbitrary platform decisions. Assessing it requires examining both effects. Equating the entire act with censorship obscures provisions that can assist a wrongly blocked user.

2.2. Systemic risk: a necessary category with room for abuse

The largest platforms and search engines must assess risks arising from their services. Categories include dissemination of illegal content, threats to fundamental rights and adverse effects on civic discourse and electoral processes. Mitigation measures must be reasonable, proportionate and effective. [7]

The difficulty begins when these categories are translated into indicators. What precisely constitutes deterioration in public debate? How should a forceful political campaign be distinguished from coordinated manipulation? How should the harm from retaining false material be compared with the harm from removing genuine reporting?

In a 2025 European Law Open article, Rachel Griffin criticises risk management for translating political conflicts into technocratic expertise and assigning substantial diagnostic responsibility to the companies being regulated. Her analysis identifies a risk of regulatory capture by powerful market participants. This is a scholarly argument about institutional design, rather than proof that every decision made under the framework is unlawful. [8]

Excessive moderation is one possible outcome of poorly aligned incentives. Another is a response confined to reports and procedures that are easy to show a regulator. Assessment should therefore ask not only how much material disappeared, but whether a specific harm declined and at what cost to lawful expression.

2.3. Trusted flaggers do not issue removal orders

Article 22 provides for priority processing of notices from entities designated as trusted flaggers. The mechanism concerns content the notifier considers illegal. Priority processing does not mean automatic removal: the platform makes the decision. [9]

The relevant national coordinator awards the status. Requirements include competence, independence from platforms, diligence and objectivity. Reporting and withdrawal of status are also provided for. Fact-checkers and trusted flaggers therefore perform distinct roles, even if an organisation may fulfil both. [9]

A privileged channel of influence still deserves scrutiny. How often are notices substantiated? How quickly are mistakes corrected? Does the organisation disclose its funding and conflicts of interest? Does it operate within its actual field of expertise? Answers should come from public evidence, rather than a general assurance that an entity is “trusted”.

2.4. The EMFA: special media safeguards and unequal access

Article 18 of the European Media Freedom Act provides particular safeguards against certain decisions by very large platforms affecting media service providers. It relies on declarations of compliance with conditions, including editorial independence and relevant standards. It is neither a general journalism licence nor a register of approved opinions. [10]

In covered situations, a media provider receives reasons for an intended restriction and generally has 24 hours to reply. This is not unconditional immunity: exceptions include actions related to specified DSA obligations and illegal content. [10]

The possibility of unequal access remains. Smaller newsrooms may find it harder to access recognised self-regulatory structures or obtain help in platform disputes. The practical availability of protection, especially for new media, should therefore be examined. This concern should not be presented as proof that the act prohibits publication by organisations outside the established market.

Commission guidelines describe Article 18 as protection against unfair moderation. They also address involvement of civil society organisations, including fact-checkers, in reviewing media declarations. This is not a general requirement to consult them over every content decision. The regulator’s stated aims and model should be tested against actual practice. [11]

2.5. Funding fact-checking: transparency instead of presumed guilt

The European Commission funds infrastructure addressing disinformation, including EDMO. In a 2026 announcement, it reported €2.5 million for the continuation of the observatory’s central activities. It also states that EDMO’s governance is designed to ensure editorial independence. Both facts should be presented together. [12]

Public funding warrants questions about transparency, but does not itself establish control over editorial judgements. Similarly, platform funding does not automatically invalidate every published check. Evidence of interference requires a specific mechanism: contractual conditions, influence over appointments, pressure over publication or a documented pattern of assessments.

An appropriate standard includes transparent grants and contracts, separation of funding from editorial decisions and a public corrections procedure. It should apply to organisations supporting regulation and to its critics alike. No one is exempt merely by describing themselves as independent.

Part III. Visibility, mistakes and the ability to challenge decisions

3.1. Shadow banning is a genuine concern that requires evidence

Shadow banning commonly describes undisclosed restrictions on an account’s or item’s visibility. It should not describe every decline in reach. Fewer impressions may result from audience interest, competition, personalisation, measurement changes or deliberate moderation. Without data, the cause cannot reliably be identified.

The central difficulty for users is the knowledge imbalance. A platform can see why a restriction occurred; its author may only see the result. The dispute should therefore concern access to information about decisions, rather than an assumption that every algorithmic decline is political punishment.

The DSA requires reasons for specified moderation restrictions. It is therefore inaccurate to say that the law necessarily leaves them outside all procedures. Problems may instead involve non-compliance, unclear explanations or difficulty distinguishing sanctions from ordinary ranking. The Commission describes statements of reasons and their database as elements of transparency. [13]

Time lost during a mistaken block cannot be fully recovered. Election reporting restored weeks later may no longer serve its purpose. This supports the case for prompt remedies, rather than proving that every decision is technically irreversible.

3.2. What do appeal statistics really tell us?

In an announcement dated 28 May 2026, covering April 2025 to March 2026, Appeals Centre Europe reported receiving more than 24,000 disputes. Where it could assess the material, it disagreed with the platform in an average of 59% of cases. The rate was 52% for disputed removals and 63% for disputed decisions to leave material up. [14]

These results reveal errors in both directions: removing content that should remain and retaining material that warrants action. They do not establish an error rate for all platform decisions. Appeals are a selected group of cases, rather than a random sample of all moderation.

Earlier reporting also requires distinguishing substantive assessment from default decisions in a user’s favour when a platform does not supply the material necessary for review. That exposes a procedural failure but is not equivalent to establishing that a particular statement complied with the rules. [15]

The claim that “three out of four removals are wrong” is therefore an unjustified generalisation. The stronger, better-supported argument is that independent appeals reveal substantial problems with decision quality and access to evidence needed for review.

3.3. Rights exist. Can citizens actually use them?

The EU framework offers channels for challenging platform decisions: internal complaints, out-of-court dispute settlement, complaints about DSA violations to the relevant coordinator and judicial protection. These are distinct procedures and do not all provide the same outcome. [16]

A form alone does not ensure effective protection. Language, cost, time, evidence submission and whether a response addresses the actual case all matter. An automated message repeating the title of a policy section may resemble an explanation while providing little help in preparing an appeal.

Assessment should measure time to decision and implementation, as well as the proportion of cases unresolved because evidence is missing. These indicators describe protection more meaningfully than the number of complaints received. Urgent communications, professional activity and material relevant to a live public debate require particular attention.

Part IV. Different countries, different mechanisms

4.1. Poland: incorrect chronology obscures the real dispute

The president refused to sign the statute of 18 December 2025 concerning electronic services on 9 January 2026, not in December 2025. The date is confirmed in the official record of presidential vetoes. [17]

The Panoptykon Foundation’s position also needs fair representation. It criticised earlier proposals, but in January 2026 opposed justifying the veto through a simplified account of unchecked censorship. Its analysis pointed to safeguards introduced and judicial challenges. Quoting only its earlier warnings while omitting its view of the subsequent text would be misleading. [18]

According to the foundation’s chronology, the president signed the first statute on 25 September 2026. It concerned national arrangements for applying the DSA. Blocking and restoration procedures were pursued separately. Signature and entry into force are different events, and neither establishes when the EU regulation itself began to apply. [19]

In a statement of 14 October 2026, the Ministry of Digital Affairs described Senate approval of a second statute concerning specified illegal content, including immediate enforcement of some decisions. This documents a particular legislative stage, rather than proving that every described provision was already in force. [20]

Poland illustrates why oversight of platforms must be distinguished from powers to intervene in individual publications. Supporting corporate accountability is compatible with demanding stronger safeguards for blocking orders. Criticism of a particular provision does not require portraying every element of the DSA as censorship.

4.2. Brazil: enforcing judgments and the cost of blocking a service

On 30 August 2024, Brazilian Supreme Court justice Alexandre de Moraes ordered X to be suspended. The court’s announcement identified compliance with orders, payment of fines and appointment of a local representative as conditions for restoration. This was therefore more than a dispute about whether posts were true. [21]

The decision also provided for a daily fine of 50,000 reais for circumventing the suspension through technological means. It was not a general ban on using VPNs for all purposes. On 8 October 2024, the court authorised the platform’s return. [21] [22]

Public authorities can enforce lawful judgments against foreign companies. A measure affecting every user, including those unrelated to violations, nevertheless needs a separate proportionality justification. That question remains valid even where the platform’s duty to comply is clear.

4.3. India and Turkey: security as a justification for restrictions

Freedom House’s Freedom on the Net 2025 report described intensified restrictions in India following the Pahalgam attack and during the conflict with Pakistan in May 2025. It identified website blocking, removal orders and arrests over online comments, alongside transparency problems in government appeal committees. These findings concern the period covered by that edition. [23]

Its Turkey report documented restricted social media access during protests, blocks affecting websites and accounts, and proceedings against journalists and users. It also recorded X restricting access to specified accounts at the authorities’ request. [24]

These cases demonstrate the risk of public-order powers being used to suppress criticism. They do not automatically establish that every national system is equivalent to the DSA. Comparisons should account for judicial independence, scrutiny of administration and opportunities to challenge decisions publicly.

4.4. Ethiopia: insufficient action can also cause harm

Restrictions on expression are not the whole platform problem. Amnesty International’s 2023 report examined Facebook’s role in human rights abuses against the Tigrayan community during the 2020–2022 conflict in northern Ethiopia. It identified the importance of the business model and serious failures in responding to dangerous content. These are a human rights organisation’s findings and should be attributed accordingly. [25]

Ethiopia challenges the simple prescription that freedom requires no moderation. Failing to respond to incitement can impede others’ exercise of their own rights. Protecting expression and preventing violence require assessment of circumstances, rather than one absolute slogan.

Nor should an entire continent be described through one country. Evidence about languages, resources and local moderation outcomes is necessary. Without it, dramatic comparisons between removal times “in Africa” and “in Europe” remain unsubstantiated generalisations.

Part V. Where does protection end and control begin?

5.1. Regulation does not eliminate the advantages of the largest firms

Reporting, audits and dispute handling cost money. It is worth investigating whether regulation entrenches companies with the largest legal teams and greatest data access. Griffin’s institutional critique raises important questions about market power. [8]

But not all businesses have identical obligations. The DSA distinguishes service types and sizes; micro and small enterprises benefit from exemptions from some requirements, while the heaviest duties concern the largest platforms and search engines. A credible cost assessment must recognise those differences. [26]

The possibility of regulatory advantage does not establish its actual extent. Comparable evidence on compliance costs, entry and exit is needed. Without it, percentages about revenue or alleged mass withdrawal from the EU market create an appearance of precision rather than proof.

5.2. Audits already exist. Their independence must be tested

Mandatory auditing should not be proposed as something entirely absent from the DSA. Very large platforms and search engines must undergo independent audits at least annually. Risk assessment and mitigation reporting are also provided for. [27]

The practical dispute concerns quality. Can auditors examine relevant data, or do they mainly rely on company documents? Do findings illuminate consequences for smaller languages and newsrooms? Do recommendations lead to change? Who discloses conflicts and checks implementation?

Demanding that auditors receive no money from either governments or companies is not enough. Every system needs funding. Transparent contracts, methodological independence, the ability to publish unfavourable findings and accountability for unreliable audits matter more.

5.3. Proportionality requires specific questions

The European freedom-of-expression standard does not prohibit every restriction. It requires a legal basis, a legitimate aim, necessity and proportionality. Public debate and journalism receive particular attention in the case law concerning Article 10 of the European Convention on Human Rights. [28]

Practical questions include whether a measure targets a specific violation or an entire service; whether a correction or context would suffice instead of removal; whether restrictions are temporary; whether prompt independent review is available; and whether affected people can see and answer the evidence.

Freedom of expression does not guarantee identical reach or automatic inclusion in recommendations. That does not relieve dominant intermediaries of responsibility for arbitrary restrictions. The more public debate depends on a single infrastructure, the more important transparent access rules become.

Part VI. An alternative: resilience without monopoly

6.1. Media literacy whose effects can be assessed

One approach is prebunking: teaching people to recognise manipulation techniques before encountering a specific false narrative. A 2022 Science Advances study by Roozenbeek and colleagues comprised seven preregistered studies, including a YouTube field experiment. The authors found improved recognition of manipulative techniques after participants watched short educational videos. [29]

This does not establish permanent immunity to every form of disinformation. The field study did not directly measure actual sharing of harmful material. Google Jigsaw’s involvement should also be disclosed. Educational interventions warrant the same scrutiny of methods and funding as moderation. [29]

A useful programme teaches source assessment, dates and context, manipulative language and the distinction between facts and interpretations. It need not require permanent trust in a single institutional list. It should teach people to say “I don’t know” and to revise conclusions when new evidence appears.

6.2. Fewer black boxes and more reader choice

Information accountability requires insight into why particular material is shown. The DSA provides for transparency concerning principal recommendation parameters and, on very large platforms, at least one recommendation option not based on profiling. [16]

The practical question is whether this choice is accessible and persistent or buried in settings. Users should understand the difference between interest-based recommendations, advertising, posts from followed authors and material promoted on other grounds. This reduces dependence on a single invisible selection logic.

Pluralism also concerns smaller newsrooms’ access to appeals, correction of mistaken labels and transparent support programmes. Journalism funding should not make agreement with the grant provider’s political position an informal condition of survival.

6.3. Six questions before restricting expression

The accountability standard proposed here can be expressed through six questions:

• What was violated? The law, a specific platform rule or simply an algorithm’s preferences?
• On what basis? What facts, evidence and criteria justify the decision?
• Who decided? The platform, a public authority or another organisation, and with what role for automation?
• What is the effect? Removal, reduced recommendations, account suspension, demonetisation or a territorial restriction?
• How can it be challenged? Where and how quickly can reconsideration and independent review be obtained?
• How can a mistake be repaired? Does correction include restoration, an explanation and mitigation of further consequences?

This is a proposed framework for assessing practice, rather than a claim that every element is already uniformly guaranteed to every user. It nevertheless shifts debate from abstract declarations of safety or freedom to verifiable conduct.

Conclusion: truth needs the possibility of challenge

The risk of a monopoly on truth does not begin only with the establishment of an office bearing that name. It can develop where the boundaries of debate are set by an opaque combination of state pressure, platform interests and expert authority, while citizens cannot determine who restricted their expression or why.

Risk should not be confused with proof of an established system. The DSA and EMFA contain regulatory instruments and safeguards for users and media. Whether they protect or undermine pluralism depends on decisions, oversight and effective challenge. Criticism that ignores existing protections is easy to dismiss. A defence of regulation that ignores possible abuse is equally inadequate.

Democracy needs protection against fraud and organised manipulation. It also needs room for error, disagreement and changing one’s mind. Institutions should not establish a final authorised account of reality. They should create conditions in which evidence can be checked, decisions challenged and mistakes repaired.

The decisive test is whether citizens can effectively challenge both false information and an unjustified decision made in the name of combating it.