Article 10 of the European Convention on Human Rights protects freedom of expression — including the freedom of the press to gather, process, and publish information without state interference. As artificial intelligence tools become integral to modern journalism, a new set of legal questions has emerged: can a state lawfully restrict AI-assisted newsgathering? Does algorithmic content moderation by platforms constitute a Convention-compatible interference? Can surveillance of journalists’ AI tools violate the confidentiality of sources? This guide analyses how Article 10 ECHR applies to AI-driven journalism, what the European Court of Human Rights has said about press freedom in the digital era, and when you can bring a successful application to Strasbourg.
Article 10 ECHR — Article 10 of the European Convention on Human Rights, which guarantees freedom of expression, including the right to hold opinions and to receive and impart information and ideas without interference by public authority. Restrictions are permissible only if they are prescribed by law, pursue one of the legitimate aims listed in Article 10(2), and are necessary in a democratic society.
Key Takeaways
- Article 10 protects journalistic method, not only content. The ECtHR has held that the right to gather information — including through digital tools — is a component of the right to impart it. Interference with AI-assisted newsgathering may engage Article 10.
- Press freedom receives the strongest protection under Article 10. States bear a heavy burden to justify restrictions on journalism. Defamation convictions, prior restraint, and orders to reveal sources require “pressing social need” and strict proportionality.
- Algorithmic filtering by state-linked platforms can constitute an “interference.” The ECtHR’s evolving case law on internet intermediaries establishes that state-compelled content removal, even via private platforms, is attributable to the state.
- Journalistic source protection extends to digital communications. Telegraaf Media Nederland v. Netherlands (2012) confirmed that surveillance measures targeting journalists to identify sources violate Articles 10 and 8. AI-powered metadata analysis raises the same risks.
- The 4-month filing deadline is absolute. Under Rule 47 of the Rules of Court, your application must reach Strasbourg within 4 months of the final effective domestic remedy. No exceptions for complexity or ongoing harm.
- Exhausting domestic remedies is mandatory. You must challenge the interference before national courts — administrative, civil, and if available, constitutional — before the ECtHR will examine your case.
What Article 10 ECHR Protects — and What It Allows States to Restrict
Article 10 paragraph 1 guarantees freedom of expression to “everyone” — individuals, journalists, publishers, broadcasters, and, in some circumstances, legal entities including news organisations. The right covers not only the final published article but the entire journalistic process: researching, gathering information, protecting sources, and deciding what to publish. The European Court of Human Rights confirmed in Dammann v. Switzerland (Application No. 77551/01) that the collection of information is a preparatory step to journalism and falls within Article 10’s scope. This is the foundation on which AI-assisted newsgathering sits: if a journalist uses an AI tool to identify document patterns, find sources, or detect disinformation, that activity is protected.
Article 10 paragraph 2 permits states to impose restrictions, but only if they satisfy a three-part test. The restriction must be (1) prescribed by law — meaning a sufficiently precise legal basis accessible to the public; (2) in pursuit of a legitimate aim listed in Article 10(2), such as national security, territorial integrity, public safety, prevention of disorder or crime, protection of reputation, or disclosure of confidential information; and (3) necessary in a democratic society — meaning there must be a pressing social need, proportionate means, and relevant and sufficient reasons. The Court has consistently held that where press freedom is at stake, the margin of appreciation left to states is narrow. In Jersild v. Denmark (Application No. 15890/89), the Court held that a television journalist convicted for broadcasting interviews with racist youths was protected under Article 10 because journalistic discretion in presentation cannot be second-guessed by domestic courts without compelling justification. Reporting on difficult subjects, even by methods that disturb public authorities, is at the core of what Article 10 protects.
Does Article 10 apply to online journalism and digital tools?
Yes, and the Court has progressively extended its Article 10 jurisprudence to digital contexts. In Delfi AS v. Estonia (Application No. 64569/09), the Grand Chamber examined whether holding a major news portal liable for third-party comments was a justified restriction. While the Court found no violation in those specific circumstances — the portal was a large commercial operator and the comments were manifestly unlawful hate speech — the judgment established important principles: states must demonstrate that liability regimes for online content are prescribed by law, pursue a legitimate aim, and do not impose a disproportionate chilling effect on news organisations. A regime that forces publishers to pre-screen all user content, for instance, would likely fail the proportionality test.
EU-funded research projects such as INJECT — a Horizon 2020 initiative involving the University of Bergen, UiT The Arctic University of Norway, Sciences Po Paris, and other partners — have developed AI tools specifically designed to assist journalists in finding angles, diversifying sources, and detecting patterns in large datasets. Such tools do not alter the legal nature of journalistic activity; they augment it. Any state measure that specifically targets the use of AI tools in journalism — through licensing, access restrictions, or forced disclosure of algorithmic parameters — would constitute an interference with Article 10 requiring full Convention justification.
AI-Assisted Journalism and Article 10: Three Legal Fault Lines
The integration of AI into newsrooms creates three areas of recurring legal tension under Article 10. First, states may attempt to regulate or prohibit the use of AI tools in journalism on grounds of national security or information integrity. Second, platforms — under state compulsion or co-regulation — may use automated systems to filter or remove journalistic content. Third, intelligence agencies may deploy AI-enhanced surveillance to identify journalistic sources through metadata analysis, behavioural profiling, or device monitoring. Each of these raises distinct Article 10 (and often Article 8) concerns.
State regulation of AI journalism tools
No ECtHR case has yet directly addressed state prohibition of AI tools used by journalists, but the Court’s general principles apply clearly. In Association Ekin v. France (Application No. 39288/98), the Court held that a blanket administrative ban on a foreign publication — with no judicial review and no proportionality analysis — violated Article 10. By analogy, a government decree barring journalists from using AI tools to cross-reference public records, without a targeted and reviewable legal basis, would face the same fate. The law must define precisely which tools are restricted, for what purpose, and with what procedural safeguards. Vague or over-broad restrictions fail the “prescribed by law” requirement.
The EU AI Act (Regulation (EU) 2024/1689), which imposes obligations on providers and deployers of AI systems, does not exempt journalistic use from its scope in all cases. High-risk AI systems used in areas affecting fundamental rights require conformity assessments. Where domestic implementation of the EU AI Act is used in ways that effectively chill journalistic AI use — by imposing prohibitive compliance burdens on small news organisations — that implementation may be challenged under Article 10 read in light of the Convention’s requirement that restrictions be proportionate and necessary.
Algorithmic content moderation and the chilling effect
Platform-level automated content moderation presents a subtler challenge. When a government compels or co-opts a platform’s AI moderation system to remove or downrank journalistic content, state responsibility may be engaged. The ECtHR in Yıldırım v. Turkey (Application No. 3111/10) found a violation when Turkish authorities blocked access to Google Sites — a hosting platform — without restricting only the offending content. The collateral suppression of lawful content, including journalistic material, was disproportionate. The same logic applies to AI moderation systems that operate as blunt instruments: if a state requires a platform to use keyword-based or classifier-based filtering that systematically flags investigative journalism as disinformation or extremist content, and the platform cannot challenge or override that requirement, Article 10 is engaged.
The chilling effect doctrine, established in Goodwin v. United Kingdom (Application No. 17488/90) and refined in subsequent press freedom cases, holds that restrictions need not result in actual suppression to violate Article 10. A credible threat of liability, de-indexing, or platform removal may suffice to deter journalistic investigation. AI moderation systems operating at scale — processing millions of items without human review — amplify this chilling effect: journalists facing automated removal have no interlocutor to appeal to, no explanation of the decision, and no rapid remedy.
Source protection against AI-powered surveillance
The protection of journalistic sources is one of the cornerstones of press freedom under Article 10. In Goodwin v. United Kingdom, the Court held that ordering a journalist to disclose sources was a disproportionate restriction unless justified by an overriding public interest. In Telegraaf Media Nederland Landelijke Media B.V. and Others v. Netherlands (Application No. 39315/06), the Court confirmed that surveillance operations targeting journalists — including interception of their communications to identify sources — violate both Article 8 (right to private life) and Article 10 (chilling effect on press freedom).
AI-powered surveillance dramatically increases the surface of source exposure. Intelligence agencies can now correlate metadata — timing of calls, location data, document access patterns — to identify sources without ever reading message content. Where domestic law permits such analysis without specific judicial authorisation linked to a concrete threat, and where the target is a journalist, Article 10 requires the Court to apply particularly strict scrutiny. In Big Brother Watch and Others v. United Kingdom (Application No. 58170/13), the Grand Chamber found that bulk interception regimes violated Article 8 in part because they lacked adequate safeguards for journalistic sources and legally privileged material. The absence of automatic notification and independent oversight were decisive. AI-enhanced bulk analysis regimes that lack equivalent safeguards are equally vulnerable.
Bringing an Article 10 Claim: What the ECtHR Requires
Article 10 claims follow the same admissibility requirements as all ECHR applications, but press freedom cases have particular procedural features. You must be a “victim” within the meaning of Article 34 — meaning you are directly affected by the measure you challenge, not merely someone who objects to a law in the abstract. A journalist whose article was removed by a state-compelled AI filter is a direct victim. A news organisation whose AI newsgathering tool was seized in a police operation is a direct victim. A journalist who self-censors because of a credible enforcement threat may also qualify as an indirect victim, but the threshold is higher and requires documentation of the chilling effect.
Domestic remedies must be exhausted. In media freedom cases, this typically means pursuing any available constitutional challenge to the restricting measure, seeking judicial review of administrative decisions (take-down orders, licensing refusals, surveillance authorisations), and if relevant, raising Article 10 expressly before national courts so the issue is on the record. The ECtHR will not re-examine factual findings made by national courts unless they are manifestly unreasonable; your domestic proceedings are where the evidentiary foundation is built. Preserve all technical evidence: access logs, content removal notices, automated moderation decisions, and any AI-generated classification outputs that affected your journalistic work.
What remedies are available if you win?
Under Article 41, the Court may award just satisfaction: pecuniary damages for financial losses caused by the interference (lost advertising revenue, legal costs of challenging take-down orders, income from suppressed publications), non-pecuniary damages for the distress and professional harm of having journalistic work suppressed, and costs and expenses of the Strasbourg proceedings. In addition, Article 46 requires states to abide by the Court’s judgments. Structural violations — such as a surveillance regime without adequate press freedom safeguards — may require the state to amend legislation under supervision by the Committee of Ministers. The Court has ordered states to introduce judicial pre-authorisation requirements for source-identification measures, amend over-broad content filtering obligations, and pay the costs of restoring removed journalistic archives.
Speed matters. The ECHR time limit of 4 months from the final domestic decision is strict and non-extendable. Where the interference is ongoing — for example, a standing court order compelling disclosure of sources or a continuing de-indexing of a publication — the clock may not start until the order is finally determined domestically, but you should take specialist advice immediately rather than assume an open-ended window exists.
Key ECtHR Cases on Press Freedom and Digital Expression
Lingens v. Austria (Application No. 9815/82, 1986). The foundational press freedom judgment. A journalist convicted of criminal defamation for publishing critical commentary about a politician was vindicated: Article 10 protects not only factual reporting but value judgments on matters of public interest. The Court held that politicians must tolerate more criticism than private individuals. This principle applies to AI-generated commentary and analysis: where an AI tool assists in producing critical opinion journalism on public affairs, defamation liability requires the same high threshold.
Observer and Guardian v. United Kingdom (Application No. 13585/88, 1991). An injunction preventing publication of extracts from the “Spycatcher” memoir violated Article 10. Prior restraint — stopping publication before it occurs — is the most serious interference with press freedom and will almost never survive Convention scrutiny. An administrative order preventing a news outlet from publishing AI-generated investigative findings before a court has reviewed them would face this same high bar.
Goodwin v. United Kingdom (Application No. 17488/90, 1996). The definitive judgment on source protection. Ordering a journalist to reveal a confidential source was not necessary in a democratic society where the leaked information had already been published and the risk of further harm was speculative. Source protection under Article 10 is not absolute — threats to life or national security can override it — but the bar is high and judicial, not administrative, authorisation is required.
Bladet Tromsø and Stensaas v. Norway (Application No. 21980/93, 1999). A regional Norwegian newspaper published allegations about seal hunters drawn from an official government inspector’s report. The Court found no violation when the paper was convicted of defamation: reliance on official documents in good faith, even if they contain errors, is protected journalistic practice. For AI journalism: where an algorithm draws on official datasets to generate factual claims, good faith reliance on those sources supports an Article 10 defence.
Magyar Helsinki Bizottság v. Hungary (Application No. 18030/11, 2016). The Grand Chamber held for the first time that access to information held by public authorities falls within the scope of Article 10 in certain circumstances — where the information is necessary for the applicant to exercise the right to freedom of expression, and where the applicant has a role as a “public watchdog.” Journalists and NGOs seeking algorithmic impact assessments or training data disclosures from state authorities may rely on this judgment to argue that refusal of access interferes with Article 10.
Delfi AS v. Estonia (Application No. 64569/09, 2015). A major news portal was held liable for manifestly unlawful comments posted by users despite prompt removal after notification. The Grand Chamber accepted that large commercial portals exercising editorial control may be treated differently from passive conduits. The case is frequently misread as license for broad platform liability; in fact, it is narrowly confined to extreme hate speech from identifiable authors on platforms with significant control over content flow. AI-driven moderation of journalistic comment sections, operating without human review, risks over-removal and may trigger Article 10 concerns in the opposite direction.
How to Apply to the ECHR in a Press Freedom Case: Practical Steps
A press freedom application to Strasbourg follows the standard application process but requires particular attention to documenting both the interference and its chilling effect. Begin by reconstructing the factual record: every communication with platforms, every administrative removal notice, every surveillance disclosure obtained through domestic proceedings. Where AI systems were involved, obtain — through data subject access requests or litigation discovery — any available documentation of the algorithmic decision that affected you. Without this material, the application will struggle to satisfy the burden of showing a concrete, not merely hypothetical, interference.
Complete the official ECHR application form (Form DOC), setting out the facts chronologically, the domestic proceedings exhausted, and the Convention rights invoked. In Article 10 cases, it is common to also plead Article 8 (private life and correspondence) where surveillance or source identification is at issue, and Article 13 (right to an effective remedy) where domestic law provides no mechanism to challenge automated content decisions. The ECHR admissibility checklist will help you assess whether your case meets the threshold requirements before you invest in full proceedings.
Speed and precision are non-negotiable. The 4-month rule runs from the date of the final effective domestic decision. Incomplete applications are rejected; missing documents cannot be submitted after the deadline. Our ECHR lawyers work with journalists, news organisations, and media NGOs facing state interference with digital journalism — from AI-driven surveillance operations to automated content removal and source disclosure orders. We advise on domestic remedies strategy, prepare Strasbourg applications, and represent clients through the full proceedings cycle.
Frequently Asked Questions
Can an AI tool be “censored” under Article 10?
A state measure that specifically prohibits or materially impedes the use of an AI tool in journalistic work — without a precise legal basis, legitimate aim, and proportionate justification — may constitute an interference with Article 10. The protection extends to the means of journalism, not only the finished product. Whether a particular restriction survives Convention scrutiny depends on the specifics: how clearly the law defines what is restricted, what aim it serves, and whether less restrictive alternatives exist.
Does Article 10 protect AI-generated journalism?
Article 10 protects persons who engage in the activity of journalism, including through AI-assisted methods. The Convention does not grant rights to AI systems themselves. Where a journalist or publisher uses AI tools to produce reporting that is then suppressed or sanctioned by the state, the human journalist or publishing entity can invoke Article 10. The use of AI tools does not reduce the level of protection — it is the journalistic purpose and the state’s interference that determine Convention liability.
Can a platform’s AI moderation system violate Article 10?
Platform conduct is not directly attributable to a state under the Convention unless the platform acts on state instructions or under a sufficiently determinative legal obligation. However, where a state compels a platform — through legislation or executive order — to deploy AI filtering systems that remove or suppress journalistic content without adequate procedural safeguards, that compulsion may engage state responsibility under Article 10. Domestic legal challenges to the underlying obligation are the appropriate first step.
How long do I have to file an Article 10 application?
4 months from the date of the final effective domestic decision. This is an absolute deadline set by Rule 47 of the Rules of Court. The clock does not pause for settlement negotiations, further domestic proceedings on related issues, or the complexity of gathering evidence. Where the interference is continuing — an ongoing content removal or an unrevoked surveillance order — the limitation period may not begin until the interference ends or a final domestic ruling is issued, but this requires careful legal analysis of the specific domestic proceedings.
What if a state denies using AI tools to surveil my journalistic sources?
States regularly deny or refuse to confirm intelligence activities. The ECtHR has accepted that in surveillance cases, applicants often cannot produce direct proof of interception. In Klass and Others v. Germany (Application No. 5029/71), the Court held that a person may claim victim status under Article 34 if they can show a reasonable likelihood of being subjected to a surveillance measure, given the legislation’s scope and their journalistic activity. Circumstantial evidence, statistical exposure, and domestic parliamentary inquiries have all been accepted as supporting an inference of surveillance. Legal representation at the domestic stage — to compel disclosure — is essential before Strasbourg proceedings.
