The two jurisdictions OffshorePress operates from were chosen separately and on different evidence. Iceland was chosen for the legislative posture set out in the Icelandic Modern Media Initiative and the country’s deliberate effort, post-Kaupthing, to make the jurisdiction a venue for journalism that other regimes would suppress. Switzerland was chosen for the older and structurally different posture set out in Article 13 of the federal constitution, the country’s distance from the Five-Eyes intelligence-sharing arrangement, and the revised Federal Act on Data Protection that came into force in September 2023. The two are not interchangeable, and the choice between them — for a customer who has decided that an offshore-hosting posture is right for the work — is worth setting out carefully.
The shorthand the operator uses internally is that Iceland is the better choice for a publication whose adversaries are commercial counterparties or copyright-leaning state actors, and Switzerland is the better choice for a publication whose adversaries are intelligence-leaning state actors. This is shorthand and it lies. The longer answer fills the rest of this article, and the longer answer also contains the cases in which the shorthand reverses or in which neither jurisdiction is the right answer at all.
The Icelandic posture
The legislative instrument that defines the Icelandic posture is the Icelandic Modern Media Initiative — the IMMI The IMMI was originally drafted by a working group convened by the Althingi’s Standing Committee on General Affairs and Education, with technical input from the Berkman Center for Internet and Society and the Reporters Sans Frontières legal-affairs office. — passed unanimously by the Althingi as Resolution 23/138 on 16 June 2010. The IMMI is not a single statute; it is a parliamentary resolution that mandated a series of statutory reforms to bring Icelandic law into alignment with the strongest free-expression and source-protection regimes documented internationally. The reforms have been implemented in stages over the subsequent decade through revisions to the Icelandic Information Act, the Media Act, the Penal Code’s libel provisions, and the Whistleblower Protection Act of 2020.
The substantive content of the IMMI reforms covers a deliberate set of doctrines. Source protection is absolute, in the sense that compelling a journalist to disclose a source requires a court ruling that no less-restrictive means exist to protect the public interest the demand asserts — a standard derived from the European Court of Human Rights’ Goodwin v. United Kingdom (1996) and applied in subsequent Strasbourg case law. Intermediary liability for hosting providers is sharply limited; an Icelandic hosting provider that takes down user content on receipt of a notice may be liable for damages to the user if the underlying content is later determined not to have been unlawful. Prior restraint — the legal instrument by which a court orders that a publication may not appear at all — is restricted to a narrow set of national-security cases and may not be used to suppress reporting on matters of public interest. Anti-SLAPP protections shift cost-recovery to the plaintiff in cases that the court determines were brought primarily to chill protected speech.
The structural consequence for a hosting customer in Iceland is that an attempt by a foreign party to compel the takedown of a customer’s workload faces, at every step, a procedural environment hostile to the demand. A US-style Section-512 notice served directly on an Icelandic hosting provider is not a binding instrument under Icelandic law; an Icelandic provider that nonetheless complied with the notice and took down the customer’s workload would be exposing itself to civil liability to the customer. A foreign court order obtained ex parte in a jurisdiction the customer has no operational connection to is not enforceable in Iceland without recognition by an Icelandic court, and the recognition standard is exacting. Mutual-legal-assistance procedures are available to foreign governments but require the foreign government to make the case in front of an Icelandic judge under Icelandic standards; the documented record of refusals is non-trivial.
Iceland is, in this frame, the closest the operator has been able to find to a civilian-press-protection regime built into the structure of the law. The country’s small population and its small legal community mean that the regime is operationally legible — a hosting customer can in fact name the relevant judges, the relevant counsel, and the relevant precedents, in the way that a customer in a larger jurisdiction with a more diffuse legal community frequently cannot. The compactness of the legal environment is part of what makes the regime work.
The Swiss posture
The Swiss posture is older, structurally different, and rests on a different set of legal instruments. The constitutional baseline is Article 13 of the Federal Constitution, which establishes telecommunications secrecy as a fundamental right and which the Swiss Federal Tribunal has applied broadly to electronic-communications infrastructure. The statutory layer is the revised Federal Act on Data Protection — the revFADP — that came into force on 1 September 2023, which establishes an EU-comparable data-subject-rights regime and which applies to the operational metadata a hosting provider holds about its customers.
The intelligence-environment context is the load-bearing element of the Swiss posture for a hosting customer whose adversaries include state intelligence services. Switzerland is not a member of the Five-Eyes intelligence-sharing arrangement, is not a “second-party” or “third-party” partner of the National Security Agency under the formulations Edward Snowden’s 2013 disclosures made public, and has historically declined to participate in the bilateral intelligence-sharing arrangements that frame the SIGINT environment for English-speaking publications. The Swiss federal intelligence service — the Nachrichtendienst des Bundes, NDB — operates under the Federal Intelligence Service Act of 2017, which requires per-target authorisation by the federal Council of Ministers for the kind of bulk collection that the Snowden-era disclosures documented as routine in Five-Eyes jurisdictions.
The federal Surveillance of Post and Telecommunications Act — BÜPF — is the statute under which Swiss law-enforcement agencies may compel a Swiss hosting provider to retain or produce customer metadata in the context of a criminal investigation. The BÜPF’s applicability to hosting providers is narrower than equivalent statutes in EU member states; the statutory definition of a “provider of derived communication services” excludes hosting providers whose service does not include real-time message delivery, and the Swiss Federal Administrative Tribunal has applied the exclusion narrowly in cases including ProtonMail’s 2019 challenge. A hosting provider whose service is virtual-machine provisioning, dedicated-machine leasing, or batch email is not, on the current case-law reading, a “provider of derived communication services” within BÜPF and is not subject to its retention or production obligations.
The legislative purpose of the Surveillance of Post and Telecommunications Act is the surveillance of communications in transit, not the retrospective production of subscription metadata held by infrastructure operators whose service does not include the delivery of communications.
The structural consequence for a Swiss hosting customer is that the legal environment offers a constitutional baseline of telecommunications secrecy, a statutory data-protection regime that is among the strongest globally outside the EU, an intelligence-environment that is materially independent of the Five-Eyes arrangement, and a surveillance statute whose applicability to hosting providers has been narrowed by case law. Swiss hosting providers may decline to comply with a foreign legal demand that has not been routed through Swiss courts, and a Swiss court applying Swiss-law standards is required to weigh the demand against the constitutional and statutory protections that the foregoing paragraphs document.
Where the postures diverge
The two postures protect different things, against different adversaries, in different procedural environments. The IMMI is a press-protection regime built around the figure of the journalist and the journalistic publication; the Swiss posture is a telecommunications-secrecy and data-protection regime built around the figure of the user and the infrastructure operator. The protections overlap but they are not identical, and the choice between them turns on what the customer’s threat model actually is.
A publication whose adversaries are commercial counterparties — a corporation whose conduct the publication has reported on, a litigation-firm acting on behalf of an offended public figure, an aggregated copyright claimant pursuing a takedown campaign — is likely to fare better in Iceland. The IMMI’s anti-SLAPP provisions, the limits on intermediary liability, and the libel-law reforms that shifted Iceland’s defamation regime away from the English common-law model and toward the European Court of Human Rights model are all directly responsive to the adversary type. A US Section-512 notice routed at the Icelandic hosting provider is, to repeat the point, not a binding instrument; the Icelandic court that hears the underlying dispute applies a free-expression standard that the US doctrine does not.
A publication whose adversaries are intelligence-leaning state actors — a foreign intelligence service whose surveillance practices the publication is reporting on, a domestic government’s signals-intelligence agency, an aligned-third-party arrangement that participates in the Five-Eyes SIGINT environment — is likely to fare better in Switzerland. The constitutional telecommunications-secrecy baseline, the documented Swiss reluctance to participate in bilateral intelligence-sharing, and the case-law narrowing of the BÜPF are all responsive to that adversary type in a way that the IMMI is not. The IMMI reforms do not address the SIGINT-environment question because the IMMI was drafted for a different problem; that is not a criticism of the IMMI, it is a statement of the regime’s scope.
A publication whose adversaries combine both categories — and the audience for OffshorePress includes publications in this position — has to decide which regime is the operationally dominant defence and place the workload accordingly. The operator’s experience is that the choice is rarely close once the customer has set out the threat model in writing; the regimes’ protections are different enough that the dominant adversary type is usually obvious in retrospect.
What the regimes do not protect against
Neither regime protects against an adversary that has the operational capability and the political will to pursue the publication outside the legal system. The IMMI does not protect against extralegal harassment of a publication’s reporters or editors. The Swiss posture does not protect against the compromise of a publication’s editorial endpoints by an adversary willing to deploy commercial-intrusion software. Neither regime protects against the publication’s own operational-security failures — a leaked password, an unencrypted backup uploaded to a cloud provider in a Five-Eyes jurisdiction, a SecureDrop instance configured against the project’s documented threat model.
Neither regime protects against the publication’s domain-name registrar, which is a separate jurisdictional decision that the hosting decision does not subsume. A publication whose hosting is in Iceland but whose domain is registered through a US-headquartered registrar with a US-jurisdiction acceptable-use policy has not made a coherent set of decisions; the registrar can be compelled by a US court order to suspend the domain, and the suspension makes the publication unreachable regardless of where the hosting actually sits. The 2017 Daily Stormer episode, in which the publication’s domain was suspended in succession by Google Domains, GoDaddy, and Tucows on the basis of acceptable-use-policy applications by the registrars, is the case study that most clearly demonstrates the registrar layer as a distinct point of failure. The publication’s politics are beside the point; the operational lesson is that the registrar can act independently of the hosting provider and that an offshore-hosting decision uncoupled from a coherent registrar-jurisdiction decision is incomplete.
Neither regime protects against the publication’s content-distribution-network provider, the publication’s TLS-certificate issuer’s certificate-revocation policy, the publication’s payment-processor’s terms of service, the publication’s email-deliverability arrangement with the major mailbox providers, or any of the other infrastructure-layer counterparties that a modern publication accumulates over the course of its operating life. The hosting decision is one decision in a stack; the rest of the stack has to be made coherently with it for the hosting decision to do the work the publication is asking it to do.
The operator’s own choice
OffshorePress operates from both jurisdictions because the audience for which the operator was set up is mixed, and because the operator considers the decision between the two regimes to be the customer’s to make rather than the operator’s. The operator’s role is to make the choice operationally available — a customer who has decided on Iceland can provision in Iceland, a customer who has decided on Switzerland can provision in Switzerland, and a customer who is undecided can read the per-jurisdiction dossiers under /jurisdictions and write to the operator’s editorial address with the questions that remain. The dossiers carry the legal-instrument citations, the carrier mix, the data-retention norms, and the surveillance-treaty status; they are the longer-form companion to this article.
The operator does not operate from a third jurisdiction at MVP and does not at present plan to add one. The reasoning is that the case for each additional jurisdiction has to be made on the same evidentiary basis as the case for the original two, and the operator has not yet found a third jurisdiction where the legislative posture, the case-law record, the carrier environment, and the operational legibility of the legal community combine to support the case. The operator considers the absence of a third jurisdiction in the catalogue more honest than the inclusion of a third jurisdiction whose posture is weaker than the first two.
A customer whose threat model is not served by either of the two jurisdictions in the OffshorePress catalogue is owed the same honest answer: the operator is not the right operator for the work, and the customer’s time is better spent reading the per-country surveys published by the Reporters Without Borders methodology team and the per-country intermediary-liability surveys published by the EFF. The catalogue is shaped to the audience the operator can serve well; a customer outside that audience deserves a referral, not a misfit subscription.