The phrase “offshore hosting” has a long tail of unflattering associations attached to it — the early 2000s file-locker boom, the late-2010s cryptocurrency-scam diaspora, and the steady commercial undertow of operators selling “DMCA-ignored” as a value proposition without the editorial scaffolding to make the slogan honest. For investigative journalism, the term carries a different weight. The structural question a newsroom faces is not whether the hosting is far away from the United States Copyright Office; it is whether the hosting sits inside a jurisdiction whose courts will defend the publication against the parties that the publication’s reporting names. This is a different conversation, and it is worth having on its own terms.
The Icelandic Modern Media Initiative, passed by the Althingi in 2010 in the wake of the WikiLeaks-Kaupthing reporting, is the cleanest legislative instance of the conversation that exists. The initiative explicitly transposes a set of source-protection, intermediary-liability, and prior-restraint doctrines from the strongest free-expression regimes of the past century into Icelandic statute, with the deliberate intent of making Iceland a venue for journalism that other jurisdictions would suppress. The point of the law is not the marketing of Iceland as a hosting destination; the point is that a publisher whose servers sit inside the jurisdiction has a substantively different legal posture against the kind of demand that takes a story down. That posture is what an offshore-hosting decision is buying for a newsroom — not anonymity, not impunity, but a defensible legal position in the venue that hears the next case.
What the threat model actually looks like
The mental model that newsrooms inherited from the open-publishing era — that legal pressure on a story is a content-neutral cease-and-desist that the publication’s lawyers will rebuff — does not match the operational reality of the past fifteen years. The pattern that has repeated, in cases ranging from the British super-injunction era through the Ecuadorian-embassy phase of the WikiLeaks saga through the more recent SLAPP-suit filings against post-Soviet anti-corruption outlets, is that the legal pressure is routed at the infrastructure layer first and the publication second. The infrastructure layer has historically been the path of least resistance: a hosting provider whose business is the thinnest of margins on a thinly-staffed customer-service desk, presented with a notice that may or may not be a binding instrument under the law of the jurisdiction in which the server actually sits, will frequently take the workload offline in order to avoid the cost of finding out.
This is the structural problem that an offshore-hosting decision is responding to. It is not that the foreign jurisdiction is more permissive of journalism; it is that the foreign jurisdiction’s hosting providers have to be served through the foreign jurisdiction’s courts, and the foreign jurisdiction’s courts apply a different — frequently more demanding — legal standard before issuing the order that an infrastructure operator must comply with. The Reporters Committee for Freedom of the Press has documented the pattern at length in its annual surveillance and source-protection guides; the Electronic Frontier Foundation’s intermediary-liability map catalogues the per-country variation that drives the choice.
The legal environment in which a publication operates includes the legal environment of its infrastructure providers. A jurisdiction that protects the press but tolerates infrastructure-layer takedowns is a jurisdiction that protects the press in form only.
The case that established the contemporary version of this problem in the public mind is the 2010 Cablegate episode, in which Amazon Web Services terminated WikiLeaks’s hosting under public-relations pressure from a US senator, days after a State Department spokesperson had explicitly disclaimed any government request for the action. The episode was not a court order; it was a commercial provider exercising its acceptable-use policy in the face of a politically charged story, and the legal record contains no court ruling that compelled the action. The lesson the episode taught the journalism-infrastructure community is that the contractual layer of the hosting relationship is a load-bearing part of the publication’s threat model, and that a hosting provider whose acceptable-use policy is pliable under political pressure is structurally unsuitable for journalism that names the parties pressuring the policy.
The jurisdictional choice is operational, not aesthetic
A hosting decision driven by the threat model the audience for this publication is working to is not satisfied by any operator that markets itself as “offshore”. The operator’s actual jurisdiction matters; the operator’s actual contractual terms matter; the operator’s actual track record under legal pressure matters. The point of choosing Iceland or Switzerland — and of choosing only those two jurisdictions — is that both have a documented legislative record of defending the press against the categories of pressure that the audience for OffshorePress works against, and both have hosting providers whose terms of service do not contain the discretionary “we may terminate at any time for any reason” clause that the US-based public-cloud operators have written into their standard contracts.
The Swiss case adds a different element to the calculation. Switzerland’s federal constitution contains an explicit telecommunications-secrecy clause (Article 13), and the country has remained outside the Five-Eyes intelligence-sharing arrangement that frames the SIGINT environment for English-speaking publications. The revised Federal Act on Data Protection (revFADP) that came into force in September 2023 adds an EU-comparable data-subject-rights regime on top of that constitutional baseline. For a newsroom whose reporting touches on the surveillance practices of governments that participate in the Five-Eyes arrangement, the operational gain from siting infrastructure in Switzerland is concrete: the legal demand from a participating government’s intelligence service must be routed through Swiss courts via a mutual-legal-assistance procedure that Swiss judges have, in the documented record, been willing to refuse on free-expression grounds.
The list of jurisdictions where a comparable case can be made is short. It is not a coincidence that the list shortens further when the analysis adds the requirement that the hosting infrastructure be operationally serviced by carriers and facilities that do not themselves carry a structural correlation to the surveillance environment of the jurisdictions whose pressure the publication is hoping to escape. OffshorePress operates from Iceland and Switzerland because the case for both jurisdictions can be made in plain prose without resort to slogan; the case for the alternatives, in the operator’s reading, cannot.
What an offshore hosting decision is not
It is worth being explicit about what siting a publication’s infrastructure in Iceland or Switzerland does not do. It does not make the publication immune to legal process from the jurisdiction in which the publication’s reporters and editors are physically located. It does not make the reporters anonymous to the parties they report on. It does not protect the publication from the structural threat that a state-aligned adversary who has compromised the publication’s editorial laptops or its source-communication channels will read the underlying material before the publication does. The infrastructure-layer decision is a layer of the threat model, not the whole of it.
The point of siting the publication’s infrastructure offshore, in the editorial register, is to make the infrastructure-layer takedown demand the most procedurally expensive element of the adversary’s available campaign. An adversary that previously could route a Section-512 notice through the publication’s US-based hosting provider and have the story offline in two business days now has to route a Mutual Legal Assistance Treaty request through the State Department, through the Icelandic Ministry of Foreign Affairs, through the Reykjavik District Court, and through the operator’s Icelandic counsel — a process that takes months and that at every step gives the publication an opportunity to litigate. The cost shift is not a guarantee of victory. It is a guarantee that the publication’s lawyers will be in court rather than negotiating with a hosting provider’s customer-service desk.
The cost shift is most visible in the cases that did not occur. The publication archives that survived the post-2013 Snowden litigation, the leak databases that survived the 2016 Panama Papers backlash, and the local-corruption archives that survived the 2018-2022 wave of European SLAPP filings are, disproportionately, the publications whose infrastructure was sited in jurisdictions whose courts the adversaries had to enter from a procedural standing-start. The pattern is not absolute and the operator does not represent it as such; the pattern is, however, persistent enough across the documented record to make the infrastructure decision worth taking seriously as part of the publication’s threat model rather than as an afterthought once the reporting is done.
The role of the operator
The operator’s role in this picture is narrow and worth being explicit about. The operator runs the layer; the operator does not edit the publication. The operator does not screen what the publication runs; the operator does not vet the publication’s reporting; the operator does not have a position on which sources the publication has cultivated or which subjects the publication has named. The operator’s editorial commitment is to the layer’s continued availability under the legal regime in which the layer sits, and to the procedural defence of the layer’s customers against demands that have not been routed through the courts of that regime.
This is a different posture from the “we agree with you” or “we’re on your side” framing that some hosting providers in the activist-infrastructure space have adopted in the past decade. The operator’s view is that the editorial work of the publication is the publication’s own and is none of the operator’s business, and that the operator’s affinity for the publication’s politics is irrelevant to the operator’s contractual obligation to defend the layer. A publication whose reporting the operator personally finds objectionable receives the same procedural defence as a publication whose reporting the operator personally agrees with. The structural reason for this posture is that the alternative — the operator picking and choosing which publications to defend on the basis of editorial agreement — would reintroduce the discretionary-termination problem that the offshore-hosting decision was meant to escape, with the operator now playing the role that the US public-cloud operators were faulted for playing.
The narrowness of the operator’s role is also the reason the operator’s catalogue is narrow. OffshorePress surfaces VPS, dedicated machines, and email service from Iceland and Switzerland, and nothing else. There is no shared-hosting tier, no Windows-RDP product, no streaming server, no GPU rental. The audience for which the operator was set up is the audience that needs the layer to hold up under the kind of scrutiny an investigative publication, a leak aggregator, or an NGO-infrastructure team draws — and the catalogue is shaped to that audience. A reader whose threat model is something else is welcome to read the catalogue carefully and decide whether the surface fits; a reader whose threat model is the audience’s is the reader for whom the catalogue is built.
What follows from the decision
The publication that decides to site its infrastructure offshore in this register has a set of operational obligations that follow from the decision. The publication’s domain-name registration is a layer that must be considered separately, since the domain registrar is its own jurisdictional decision and an adversary who can compel the registrar can make the publication unreachable regardless of where the hosting sits. The publication’s source-communication infrastructure — its SecureDrop instance, its OnionShare-mediated drops, its Signal account — sits at a layer that the hosting decision does not protect. The publication’s editorial laptops, the physical security of its newsroom, the operational discipline of its reporters when travelling, all sit at layers the hosting decision does not address.
The hosting decision is one decision in a stack of decisions. The publications whose infrastructure has held up under the kind of pressure the audience for this article works against are the publications that have made the entire stack of decisions in coherence with one another. The operator’s view, defended at greater length in the OffshorePress operating principles, is that a coherent stack of decisions is what an honest hosting customer is buying when they read past the marketing to the contract — and that the publication of the operator’s own posture, in the publication, is the precondition for the customer to make that judgement on the strength of evidence rather than slogan.
The reporting that has held up against the legal pressure of the past decade is the reporting whose publishers thought about all of these layers in advance and operated accordingly. The reporting that has not is the reporting whose publishers thought the adversary would knock at the front door, and were surprised to learn that the adversary’s first move was a procedurally efficient takedown notice routed through a hosting provider whose customer-service desk had no incentive to litigate on the publication’s behalf. The point of an offshore-hosting decision is to be in the first category rather than the second. The point of writing this in the journal of the operator that runs the offshore hosting is that an audience reading carefully, from inside its own threat model, deserves to know what the decision is for and what the decision is not for, in plain prose, before any commercial relationship with the operator is contemplated.
The catalogue is one click away in the navigation bar at the head of this page. The reader who has weighed the argument set out here against their own threat model is the reader for whom the catalogue exists. The reader who has not yet weighed the argument is welcome to spend more time in the journal first; the rest of the founding set of articles is the operator’s attempt to set the rest of the operating posture out at the same length and in the same register as the argument that begins here.