Skip to main content
Thierry Gilgen

Edition history

Reading
  1. 4
    Edition 4

    Published content updated

  2. 3
    Edition 3

    Published content updated

  3. 2
    Edition 2

    Published content updated

  4. 1
    Edition 1

    Published content updated

The Website That Was Never Seized · Edition 2

SHA-256: a24f051a7d0e9d6fe9010760d93c384db2f21951fddc86f5895f482df959561e

## The short version
The Swiss Federal Supreme Court did not declare that the internet should be beyond law. It decided something simultaneously narrower and more demanding: the general Swiss rules governing seizure and confiscation were not a sufficient legal basis to compel an internet access provider to manipulate its DNS responses. The judgment, **7B_52/2026**, was issued on **12 October 2026** and made public on **4 November**. It arose from allegations that an online investment operation caused more than **CHF 1.7 million** in losses to dozens of people. Ten providers had been named in a July 2025 order concerning twelve web addresses. One provider fought the order to the highest court.[^1][^2]
The same legal technique had also been used in a different case involving the sites of a climate activist movement. There, a Vaud court had already overturned a blocking order because the necessary connection between the websites and the suspected offence was not sufficiently established. Meanwhile, a court in Valais had accepted a more expansive interpretation of seizure law in principle, even while rejecting a technically unworkable order. The problem was not a single bureaucratic error. It was a live disagreement about the powers of the state, the meaning of an internet address, and who must do what when the law commands a machine to make something disappear.[^3][^4][^5]
Then, days after the Supreme Court ruling became public, a document appeared in a LinkedIn post. Dated **9 November 2026**, it appeared to show a Vaud prosecutor lifting the seizure of two redacted web addresses in a separate proceeding and citing that new judgment. That document is a powerful epilogue **if authenticated**. It is not part of the Supreme Court case record, and the image alone cannot prove that every ISP had removed every block.[^6]
The question at the heart of this story is not whether fraud should be stopped. It should. The question is what precisely a government is authorised to do to an infrastructure it does not own, when the mechanism available to it works differently from the object described in its legal order.
The Swiss Federal Supreme Court judgment **7B_52/2026** was issued on **12 October 2026** and made public on **4 November**. It held that the general Swiss rules on seizure and confiscation were not a sufficient legal basis to compel an internet access provider to manipulate its DNS responses. The case arose from allegations that an online investment operation caused more than **CHF 1.7 million** in losses to dozens of people. Ten providers had been named in a July 2025 order concerning twelve web addresses. One provider fought the order to the highest court.[^1][^2]
The same legal technique had also been used in a different case involving the sites of a climate activist movement. There, a Vaud court had already overturned a blocking order because the necessary connection between the websites and the suspected offence was not sufficiently established. A court in Valais had accepted a more expansive interpretation of seizure law in principle, even while rejecting a technically unworkable order.[^3][^4][^5]
Days after the Supreme Court ruling became public, a document appeared in a LinkedIn post. Dated **9 November 2026**, it appeared to show a Vaud prosecutor lifting the seizure of two redacted web addresses in a separate proceeding and citing that new judgment. That document would be a significant follow-up **if authenticated**. It is not part of the Supreme Court case record, and the image alone cannot prove that every ISP had removed every block.[^6]
The question is what a government is authorised to do to an infrastructure it does not own, when the mechanism available to it works differently from the object described in its legal order.
## I. Twelve addresses
On **15 July 2025**, an order from the public prosecutor in the canton of Vaud set out to stop a particular kind of harm.
Investigators had been pursuing an alleged online investment fraud since **22 August 2024**. The accusation described a familiar deception: an apparently legitimate platform for stock-market investments, people with limited experience of trading, payments they believed were being invested, and money allegedly appropriated instead. The record placed the losses above **CHF 1.7 million** and referred to dozens of victims. These were accusations in an ongoing criminal investigation, not findings of guilt against named individuals.[^1]
The prosecutor's July order named **ten internet companies** and listed **twelve web addresses** to be seized with immediate effect. It invoked the possibility of criminal fines under Article 292 of the Swiss Criminal Code for non-compliance. The order reached beyond the suspected operators of the fraudulent service. It reached companies that carried other people's traffic.[^1]
This is the part of the story that becomes lost when a dispute is reduced to *free internet versus censorship*. People had reportedly lost serious money. If another person was about to send savings to the same criminals, intervention might protect them. The interest was concrete, not theoretical.
There is, however, a problem with issuing instructions to the wrong part of a system.
An internet access provider does not necessarily operate the suspicious website. It may not host the server, register the domain, hold the customer account associated with it, or possess the credentials that could remove its content. It can carry the connection between an end user and the wider internet. It can often influence the answers given by the DNS resolver it supplies. Those are very different forms of control.
To a non-technical reader, *seize the website* might sound like one sensible instruction. To an engineer, that phrase is incomplete. Seize the server? Suspend the hosting account? Change the registrar's registration record? Withdraw a domain from a registry zone? Intercept traffic? Change what a resolver says? Each operates at a different place in the chain and has different collateral effects.
The Vaud dispute turned, in a surprisingly literal way, on what a state meant when it said it wanted to seize a web address.
The prosecutor's July order named **ten internet companies** and listed **twelve web addresses** to be seized with immediate effect. It invoked the possibility of criminal fines under Article 292 of the Swiss Criminal Code for non-compliance. The order reached beyond the suspected operators of the fraudulent service, to companies that carried other people's traffic.[^1]
People had reportedly lost serious money. If another person was about to send savings to the same criminals, intervention might protect them.
An internet access provider does not necessarily operate the suspicious website. It may not host the server, register the domain, hold the customer account associated with it, or possess the credentials that could remove its content. It can carry the connection between an end user and the wider internet. It can often influence the answers given by the DNS resolver it supplies. Those are different forms of control.
*Seize the website* is therefore an incomplete instruction. Seize the server? Suspend the hosting account? Change the registrar's registration record? Withdraw a domain from a registry zone? Intercept traffic? Change what a resolver says? Each operates at a different place in the chain and has different collateral effects. The Vaud dispute turned on which of these the state was in fact ordering.
<figure> <img src="/api/media/field-notes/website-that-was-never-seized-f01-en-fdd3ae106d87e3a525143000d265441df25eef316882026698d423d83dcdc062.webp" alt="Four separate metric cards: alleged losses above CHF 1.7 million, ten providers named, twelve web addresses listed, and five blocking orders reported by Init7 lawyers." /> <figcaption>Figure 1. Four distinct measures, not four compatible units. The first three come from the Supreme Court record; the five-order count was reported by lawyers representing Init7. The CHF amount describes alleged loss, not adjudicated damages.</figcaption> </figure>
The government cannot simply choose a useful verb and assume the infrastructure knows which lawful action follows.
It must identify the act it is ordering, the legal power that authorises it, the party capable of performing it, and the people who will be affected.
## II. The machine's address book
Imagine a building with a public address and several rooms inside it. An investigator believes that an illegal transaction is taking place in one room. The front desk has a list matching the street address to the building, but no floor plan and no knowledge of what happens behind each door.
Now order the front desk to make one room vanish from that address book.
The problem is not one of goodwill. The book does not contain room numbers.
DNS works at a comparable level of abstraction. It is part of the system that allows users to ask for an internet hostname and receive information needed to find the associated service. A browser accessing `https://example.org/offer` may need a DNS answer concerning `example.org`. The path `/offer` does not form part of the DNS question. The DNS resolver cannot decide, based on that query alone, whether the user is seeking `/offer`, `/press`, `/login`, or another path on the same hostname.[^7]
A building has a public address and several rooms. An investigator believes that an illegal transaction is taking place in one room. The front desk has a list matching the street address to the building, but no floor plan. The address book does not contain room numbers, so the desk cannot make one room vanish from it.
DNS works at a comparable level. It allows users to ask for an internet hostname and receive information needed to find the associated service. A browser accessing `https://example.org/offer` may need a DNS answer concerning `example.org`. The path `/offer` does not form part of the DNS question. The resolver cannot decide, from that query alone, whether the user is seeking `/offer`, `/press`, `/login`, or another path on the same hostname.[^7]
<figure> <img src="/api/media/field-notes/website-that-was-never-seized-f02-en-1b25aa600e71f61ded1070ca6db00b79f511a27732d1f846a02c93e2be1243c0.webp" alt="Diagram of two paths on one hostname, a single DNS hostname lookup, and a domain-level block that can affect both paths." /> <figcaption>Figure 2. Illustrative DNS decision. One hostname can serve lawful and unlawful paths. A resolver-level block operates at the hostname resolution layer; the exact scope depends on the configuration of the order and the block.</figcaption> </figure>
A DNS-based measure commonly substitutes an error, no answer, or another answer when customers of a particular resolver ask for a blocked name. It can be effective as friction. Someone who types the website into a browser, trusts their provider's default settings, and sees an error may stop there. That can matter to a person about to become a fraud victim.
But friction is not possession.
The underlying server is not taken over. A domain registered elsewhere is not confiscated. The content is not automatically deleted. A customer using an independent resolver, a different access network, or another route may still reach it. The precise outcome depends on additional measures, protocol details, client configuration, caches, and network policy.[^1][^3]
DNS also knows surprisingly little about what a human user thinks they are asking for. Blocking the hostname can block innocent pages on that hostname. Depending on exactly which names are filtered, services such as mail routing, application endpoints or other subdomains can also be affected. One should not claim that a block of a single hostname inevitably disables all email under the registered domain. The correct claim is more careful: a broad domain-level DNS restriction may create collateral effects beyond the targeted URL; the precise effects are configuration-specific.[^7]
A DNS-based measure commonly substitutes an error, no answer, or another answer when customers of a particular resolver ask for a blocked name. That friction can be effective. Someone who types the website into a browser, trusts their provider's default settings, and sees an error may stop there. That can matter to a person about to become a fraud victim.
It does not, however, take possession of the site. The underlying server is not taken over. A domain registered elsewhere is not confiscated. The content is not automatically deleted. A customer using an independent resolver, a different access network, or another route may still reach it. The precise outcome depends on additional measures, protocol details, client configuration, caches, and network policy.[^1][^3]
Blocking the hostname can also block innocent pages on that hostname. Depending on exactly which names are filtered, services such as mail routing, application endpoints or other subdomains can be affected. A block of a single hostname does not inevitably disable all email under the registered domain. A broad domain-level DNS restriction may create collateral effects beyond the targeted URL; the precise effects are configuration-specific.[^7]
The technical expert opinion from **Switch**, dated **23 February 2026** and later published alongside reporting on the dispute, was central to explaining this mismatch. Switch is not an internet-access provider in the same sense as the companies ordered to change customer-facing resolver responses. Its expertise lies in naming infrastructure, among other activities. The opinion distinguished a URL from the domain information available to DNS and warned about overblocking.[^7][^3]
This is a useful moment for the reader to notice how small the disagreement can appear in ordinary language. A prosecutor says *URL*. A network operator hears *hostname*. A judge hears *seizure*. The three terms are not synonyms, and the network cannot make them interchangeable by judicial enthusiasm.
A sophisticated technical system can be obedient to the wrong command just as easily as to the right one.
The first condition of legitimate control is knowing what is actually controlled.
In ordinary language the disagreement can look small. A prosecutor says *URL*. A network operator hears *hostname*. A judge hears *seizure*. The three terms are not synonyms.
## III. Ten networks, different answers
The original ruling identifies the provider that appealed as **A.________ AG**, the anonymised party in the published judgment. Contemporary reporting identifies the litigating company as **Init7**, the independent Winterthur-based internet access provider. Its CEO, Fredy Künzler, became a public face of the dispute.[^1][^2][^8]
The company did not merely object in a blog post. It challenged the orders. Its lawyers argued that seizure and DNS blocking were legally and technically distinct, and that the general Swiss seizure provisions did not provide a sufficiently explicit foundation for forcing an access provider to suppress its customers' DNS lookups.[^3][^5]
Init7 challenged the orders. Its lawyers argued that seizure and DNS blocking were legally and technically distinct, and that the general Swiss seizure provisions did not provide a sufficiently explicit foundation for forcing an access provider to suppress its customers' DNS lookups.[^3][^5]
According to reporting based on those lawyers' accounts, Init7 was dealing with **five** blocking orders across Vaud and Valais. The disputes involved fraudulent websites, a climate movement's sites, and further orders. One first order seeking seizure of domain names had been superseded after Init7 pointed out that it was an access provider, not the registrar capable of seizing the names. The authority then interpreted its desired outcome as access blocking instead.[^4]
The scale matters, but not in the way a chart of providers might suggest. The **ten companies** in the July 2025 Supreme Court case and the **five orders** reported across the wider dispute are not units from the same dataset. They should not be plotted as if one were a percentage of the other.
Reporting says larger providers including Swisscom, Sunrise and Salt implemented blocking measures while Init7 resisted. Those are reported actions, not a judicial finding that the companies shared the same legal interpretation, intended to censor speech, or knowingly broke net neutrality. An order backed by sanctions changes how a company evaluates its options. A cautious article has to recognise both the pressure and the option of judicial review.[^3][^8][^9]
The pressure was not merely commercial. Reporting described a **CHF 6,000 sanction against Künzler**, contested at the time. The reporting available for this draft does not establish its final resolution by 11 November 2026; its outcome should not be inferred from Init7's later victory on the DNS issue.[^4]
A company can reasonably argue that police and prosecutors need functioning tools against online fraud. Another can reasonably argue that providers should not be forced to engineer legally questionable interference with access to lawful information. The difficult question is who decides, when, and under which review process.
Init7 made that question unavoidable.
What gives the company's resistance broader relevance is not a romantic idea that small businesses are always virtuous or large ones always compliant. It is that infrastructure companies are often the point at which government orders become customer-visible facts. The law may be abstract, but the altered resolver answer is operational.
There is no need to assign saints and villains to understand the institutional significance.
A network operator found an order that it believed exceeded the state's authority. It carried the dispute into court. That is a measurable exercise of legal recourse.
The **ten companies** in the July 2025 Supreme Court case and the **five orders** reported across the wider dispute are not units from the same dataset. They should not be plotted as if one were a percentage of the other.
Reporting says larger providers including Swisscom, Sunrise and Salt implemented blocking measures while Init7 resisted. Those are reported actions, not a judicial finding that the companies shared the same legal interpretation, intended to censor speech, or knowingly broke net neutrality. An order backed by sanctions changes how a company evaluates its options.[^3][^8][^9]
Reporting also described a **CHF 6,000 sanction against Künzler**, contested at the time. The reporting available for this edition does not establish its final resolution by 11 November 2026. That outcome should not be inferred from Init7's later victory on the DNS issue.[^4]
## IV. How a seizure became a block
The cantonal judges in Vaud rejected Init7's appeal on **16 December 2025**. The company lodged a further appeal with the Federal Supreme Court on **14 January 2026**. Its request to suspend enforcement while the appeal proceeded was rejected on **3 February**. That procedural history reveals a problem that technical specialists often experience long before a final judgment arrives: legal review takes time, while a blocking order can have immediate effects.[^1]
The prosecutor's legal argument was not frivolous. It rested on a functional analogy.
Swiss criminal procedure allows provisional seizure of certain objects and assets where they may ultimately be confiscated. The prosecutor argued that a DNS block was less intrusive than physically taking away information-technology equipment, and that the ability to impose the greater intervention should encompass the lesser. The canton had viewed the distinction between seizure and DNS blocking as excessively formalistic for a technological world.[^1]
Think through the appeal of this reasoning. A server could, in principle, be seized if the law allows it and the necessary jurisdictional conditions exist. Removing an entire server can affect other users and businesses. Asking providers to make a fraud site hard to reach might appear less disruptive. If the objective is to prevent harm, why should a less invasive instrument require an additional statute?
Because an instrument is not legally smaller merely because it looks technically simpler.
The Supreme Court distinguished confiscation, which confers a form of control over the object being taken, from DNS blocking, which imposes a new active obligation on third-party internet access providers. The online service remains available through other networks or resolvers. The effects fall on providers and on the public's experience of access, not just on the person suspected of using the site criminally. This was not merely a debate about terminology; it was about the legal identity of an act.[^1]
The cantonal judges in Vaud rejected Init7's appeal on **16 December 2025**. The company lodged a further appeal with the Federal Supreme Court on **14 January 2026**. Its request to suspend enforcement while the appeal proceeded was rejected on **3 February**. Legal review takes time; a blocking order can have immediate effects.[^1]
The prosecutor's argument rested on a functional analogy. Swiss criminal procedure allows provisional seizure of certain objects and assets where they may ultimately be confiscated. The prosecutor argued that a DNS block was less intrusive than physically taking away information-technology equipment, and that the ability to impose the greater intervention should encompass the lesser. The canton had viewed the distinction between seizure and DNS blocking as excessively formalistic for a technological world.[^1]
A server could, in principle, be seized if the law allows it and the necessary jurisdictional conditions exist. Removing an entire server can affect other users and businesses. Asking providers to make a fraud site hard to reach might appear less disruptive. The Supreme Court did not treat the technically simpler-looking instrument as legally smaller for that reason.
It distinguished confiscation, which confers a form of control over the object being taken, from DNS blocking, which imposes a new active obligation on third-party internet access providers. The online service remains available through other networks or resolvers. The effects fall on providers and on the public's experience of access, and on the person suspected of using the site criminally.[^1]
The judges then considered the structure of Swiss legislation. Parliament had expressly created blocking or domain-level powers in specific settings. The Gambling Act contains a regime for unauthorised online gambling offers, including designated authorities, published lists, objections and compensation arrangements. Telecommunications law contains a specific mechanism for particular categories of prohibited sexual material involving minors. Under unfair-competition legislation, Swiss-managed second-level domains can be subject to a different registry-facing process.[^1]
<figure> <img src="/api/media/field-notes/website-that-was-never-seized-f03-en-ddc08f6a47e34d201116df75adad7c6b1ef422f996587a3c00973387dc59843f.webp" alt="Matrix of the rejected general seizure mechanism, three existing specialised regimes, and the separate draft Article 6b, marked as a consultation draft not enacted." /> <figcaption>Figure 3. Legal pathways as of 11 November 2026. “Express statutory regime” does not mean unlimited blocking authority. Each mechanism has different conditions. The draft Article 6b is a proposal, not current law.</figcaption> </figure>
If those special provisions exist, why would the broad seizure clauses silently authorise the same technique in any ordinary criminal case? That was a substantial part of the Supreme Court's reasoning. The court was not deciding that a DNS block could never prevent crime. It acknowledged that this could be a useful instrument, especially against phishing or investment fraud. It decided that the choice of offences, authorities, procedures and limits belonged primarily to the federal legislature.[^1]
The result was unequivocal for the issue before it: **Articles 263(1)(d) CPP/StPO and 69(1) CP/StGB did not authorise an order to an internet access provider to implement a DNS block.** The contested order against the appellant was set aside with immediate effect.[^1]
That is different from declaring all website blocking illegal. It is also different from determining that no domain could ever qualify as an object of confiscation. The court explicitly left that broader domain-as-object question unresolved because the case concerned resolver manipulation by a third party.[^1]
There are two important lessons in this distinction. First, the wording of a statute must be read together with its purpose, system and consequences. Second, a public authority cannot treat the absence of a tailored rule as an invitation to manufacture that rule through an expansive reading of general coercive powers.
A computer network may make a new form of action possible. It does not, by itself, make that action authorised.
A substantial part of the Supreme Court's reasoning was that those special provisions made it hard to treat the broad seizure clauses as silently authorising the same technique in any ordinary criminal case. The court acknowledged that a DNS block could be a useful instrument, especially against phishing or investment fraud. It decided that the choice of offences, authorities, procedures and limits belonged primarily to the federal legislature.[^1]
The result for the issue before it was unequivocal: **Articles 263(1)(d) CPP/StPO and 69(1) CP/StGB did not authorise an order to an internet access provider to implement a DNS block.** The contested order against the appellant was set aside with immediate effect.[^1]
The court did not declare all website blocking illegal. It also left unresolved whether a domain could ever qualify as an object of confiscation, because this case concerned resolver manipulation by a third party.[^1]
## V. The site about climate
The fraud case might have stayed a specialist dispute about criminal procedure, telecommunications and the stubbornness of one provider. The climate-group case exposed a different side of the same machinery.
The movement **Grondements des Terres** came under investigation in connection with a suspected arson attack on a Holcim cement facility in Bretonnières in June 2025. Reporting described material damage at the facility and a request that providers block sites associated with the movement. The reporting also said that prosecutors had not identified an explicit criminal call on the website itself; the authorities' reasoning involved the group's online presence, including a reference to its Instagram account.[^3]
Those statements must be kept separate. The attack is an event. Suspicion that an organisation or particular individuals were connected to it is an allegation. The contents of a website are a further evidentiary question. A blocking order is still another question, requiring its own justification.
To say that the website was associated with activists is not to say its content was illegal. To say a website included a link to another social profile is not to show that it helped commit a crime. Those are steps in a chain that cannot be skipped.
In **May 2026**, the Vaud cantonal court annulled the order against the group's websites. According to reporting of the decision and the lawyer's explanation, the necessary connection between the websites and the suspected offences was not even sufficiently plausible. The court did not settle the sweeping question of whether every DNS blocking order under the general seizure legislation was impermissible. It resolved the narrower matter: this particular factual basis was not enough.[^4]
Why does that matter in a volume principally about the Supreme Court's later ruling?
Because the alleged harm that motivates a power is not a reliable limit on every future use of that power.
A technique introduced to make investment fraud more difficult can be directed at a politically controversial website. The justification may be entirely different. The evidentiary connection may be much weaker. Yet the mechanism presented to an ISP is the same: stop people from reaching this address through your resolver.
That does **not** prove an intentional political censorship campaign. It does show how the existence of a technically convenient enforcement lever creates pressure to apply it across categories of cases. If the legal rules are vague, the point at which a prosecutor's suspicion becomes a network-wide restriction can become dangerously uncertain.
The right question is not whether someone agrees with the activists, dislikes their tactics, supports climate action, or condemns the underlying attack. None of those positions decides whether this website should have been blocked.
The right question is what evidence and legal authority must exist before the public's access to a speaker's site is restricted.
Democratic freedoms are tested most usefully at the point where protecting them is uncomfortable.
Those points have to be kept separate. The attack is an event. Suspicion that an organisation or particular individuals were connected to it is an allegation. The contents of a website are a further evidentiary question. A blocking order requires its own justification. Association with activists does not make the site's content illegal. A link to another social profile does not, by itself, show that the site helped commit a crime.
In **May 2026**, the Vaud cantonal court annulled the order against the group's websites. According to reporting of the decision and the lawyer's explanation, the necessary connection between the websites and the suspected offences was not even sufficiently plausible. The court did not decide whether every DNS blocking order under the general seizure legislation was impermissible. It held that this particular factual basis was not enough.[^4]
A technique introduced to make investment fraud more difficult can be directed at a politically controversial website. The justification may be different, and the evidentiary connection much weaker, while the mechanism presented to an ISP remains the same: stop people from reaching this address through your resolver. That fact does not establish an intentional political censorship campaign. It does show how a technically convenient enforcement lever can be applied across categories of cases. If the legal rules are vague, the point at which a prosecutor's suspicion becomes a network-wide restriction is hard to fix in advance.
<figure> <img src="/api/media/field-notes/website-that-was-never-seized-f04-en-c2410a3854ec7cf38b0bd8ac31892089946d25dde383bd18e4ca75008486083f.webp" alt="Conceptual editorial illustration of sealed Swiss documents and a barrier in a glowing network landscape, with paths continuing beyond the barrier. AI-generated, not a court record or a photograph." /> <figcaption>Original editorial illustration. AI-generated conceptual art, not a photograph, a courtroom reconstruction, or a reproduction of a prosecutor's order. The redacted document is symbolic.</figcaption> </figure>
## VI. One country, different readings
The most revealing evidence that this was a serious legal dispute rather than a single absurd instruction comes from **Valais**.
A separate fraud investigation there began after a complaint in June 2025. On **27 October 2025**, authorities issued seizure and blocking orders directed at providers. The cantonal court considered the company's appeal in decision **P3 25 268** of **28 May 2026**.[^5]
A separate fraud investigation in **Valais** began after a complaint in June 2025. On **27 October 2025**, authorities issued seizure and blocking orders directed at providers. The cantonal court considered the company's appeal in decision **P3 25 268** of **28 May 2026**.[^5]
The Valais judges accepted, as a matter of principle, a broader reading of the power to seize internet domains or interrupt access. They reasoned that the virtual character of a domain was not necessarily an obstacle to the criminal-seizure framework, drawing comparisons with freezing bank assets and with other legal approaches to digital objects. At the same time, they concluded that the particular order's technical parameters were not workable as drafted, so the order could not stand in that form.[^5]
It was a split result, and that is precisely what makes it historically interesting.
It shows two courts answering different questions at different levels. The Valais court saw an adequate conceptual bridge between seizure law and access restriction but rejected the implementation. The later Supreme Court saw the bridge itself as legally inadequate for an order imposed on an ISP's DNS resolvers under the general provisions. An engineer might have applauded the Valais technical criticism while worrying about the legal principle. A prosecutor might have applauded the legal principle while finding the decision operationally frustrating.
Neither position can be properly understood by extracting a single victorious sentence.
The Supreme Court also considered legal scholarship that had reached genuinely different conclusions. Some authors found the general provisions inadequate. Others saw a way to permit DNS blocking as a substitute for seizure in a networked environment, especially when the suspicious domain existed abroad and could not be taken over through Swiss registries. The judgment discussed those arguments rather than pretending everyone had always agreed.[^1]
This matters for how a history of the case should be written. A documentary should show the strongest case for the prosecutor's desired outcome, not only the weaknesses exposed after the fact. It should explain why law often meets technology by analogy and why analogies can fail when the technical object behaves differently from a physical asset.
The highest court did not say the policy was impossible. It said that the policy choice had to be made transparently by lawmakers.
The distinction between an inconvenient law and a law that does not yet exist is not a technicality. It is what separates interpretation from legislation.
The later Supreme Court, by contrast, held that the general provisions were legally inadequate for an order imposed on an ISP's DNS resolvers.
The Supreme Court also considered legal scholarship that had reached different conclusions. Some authors found the general provisions inadequate. Others saw a way to permit DNS blocking as a substitute for seizure in a networked environment, especially when the suspicious domain existed abroad and could not be taken over through Swiss registries. The judgment discussed those arguments.[^1]
The highest court did not say that a blocking policy was impossible. It said that the policy choice had to be made transparently by lawmakers.
<figure> <img src="/api/media/field-notes/website-that-was-never-seized-f05-en-52a8983658de7161d5f667131fb45711f8f8b1794a29c224100745baa91b3aad.webp" alt="Four-row chart distinguishing the main Vaud fraud case, the climate-website case, the Valais ruling, and the 9 November screenshot-based follow-on order." /> <figcaption>Figure 4. Four different procedural records. The 9 November row derives from a screenshot, not an independently obtained original. These are not four judgments reaching the same holding.</figcaption> </figure>
## VII. The network is not neutral about power
The expression *net neutrality* often enters this conversation too quickly.
Swiss telecommunications law sets out a principle of an open internet in Article **12e FMG**. As BAKOM has explained in its public commentary, this principle has been part of Swiss federal telecom law since 2021.[^10] Its application has qualifications, including circumstances involving the implementation of law or judicial decisions. Whether particular providers independently breached that provision in obeying the Vaud or Valais orders is **not settled by judgment 7B_52/2026**.[^11][^1]
A company that receives an official instruction, on pain of possible punishment, is not in the same position as one that invents a blocking policy on its own. Nor is compliance with an order proof that the order was legally valid. Both propositions can be true at once.
The Supreme Court ruling adds something important to the debate about infrastructure power. It shows that the practical freedom to operate an open network depends on institutions capable of reviewing the rules imposed on it. A provider may not be able to litigate every doubtful instruction. A government may have serious reasons to act urgently. The choice of who carries the burden of challenge, how appeals work, and what happens before an appeal is heard shapes the real experience of freedom.
In the July order, ten internet operators were told to act. Even if they were equal before the law, they were not necessarily equally willing, resourced or equipped to challenge the measure. That asymmetry can have visible consequences. Some users may receive ordinary DNS answers. Others may receive blocked responses. A visitor who cannot reach a site may not know whether the problem is technical failure, private filtering, an official court decision, or a prosecutor's contested instruction.
The public experiences a result; the machinery of authority is often invisible.
This is why a discussion of digital sovereignty that stops at server location is inadequate. The service may sit on a machine in Switzerland. Its domain may be registered abroad. Its public DNS may be run by a third party. Its customers may use resolvers controlled by large access providers. The decision to block it may come from a different canton. The authority to review that decision may sit in Lausanne. Sovereignty in such a system is not a colour on a hosting map.
It is a stack of jurisdiction, technical control, accountability and remedy.
A company that receives an official instruction, on pain of possible punishment, is not in the same position as one that invents a blocking policy on its own. Compliance with an order is also not proof that the order was legally valid. Both can be true at once.
In the July order, ten internet operators were told to act. Even if they were equal before the law, they were not necessarily equally willing, resourced or equipped to challenge the measure. Some users may then receive ordinary DNS answers, and others blocked responses. A visitor who cannot reach a site may not know whether the problem is technical failure, private filtering, an official court decision, or a prosecutor's contested instruction.
The service may sit on a machine in Switzerland. Its domain may be registered abroad. Its public DNS may be run by a third party. Its customers may use resolvers controlled by large access providers. The decision to block it may come from a different canton. The authority to review that decision may sit in Lausanne. Jurisdiction, technical control and the available remedy can therefore sit in different places.
<figure> <img src="/api/media/field-notes/website-that-was-never-seized-f06-en-a178b6ede8819e715401a1b793dfd73cfc19c1bee37d47e43b79e3483f74a67e.webp" alt="Side-by-side map of four technical control planes: access-provider resolvers, domain registries, hosting providers, and end-user or browser choices." /> <figcaption>Figure 5. A conceptual control map, not a catalogue of Swiss legal permissions. A DNS resolver, a domain registry, a hosting company and an end-user browser have different powers and produce different effects.</figcaption> </figure>
A mature society should be able to ask four questions every time a technology becomes an enforcement tool. Who is actually capable of performing the action? What other people or services will be affected? What law authorises the instruction? How can an erroneous order be detected, reversed and audited?
Without those answers, *sovereign control* can become a euphemism for unreviewable technical pressure.
Four practical questions follow. Who is actually capable of performing the action? What other people or services will be affected? What law authorises the instruction? How can an erroneous order be detected, reversed and audited?
## VIII. A piece of paper, nine November
The Supreme Court issued its decision on **12 October 2026**. It became publicly available on **4 November**. On **9 November**, a document appeared in a social-media post that suggests the judgment was already influencing a different prosecutorial file.[^1][^6]
The document in the screenshot has the visual form of a decision from the **Ministère public central**, *Division affaires spéciales*, canton of Vaud. It carries the file reference **PE25.026767-JDZ**, not the **PE24.017098** case number associated with the Supreme Court decision. It refers to an original seizure order dated **25 February 2026**. It shows two redacted domain names, one ending in `.com`, another in `.org`. The decision date reads **9 November 2026**. It invokes the Federal Supreme Court ruling **7B_52/2026** and states that the URL seizure must be lifted.[^6]
The visual evidence supports this reading of what the screenshot purports to show. It does not establish who first published the order, whether the image is complete, whether every named recipient received it, or whether corresponding DNS filters were subsequently removed. Nor should the identity of the accused person be inferred from the redactions. The underlying offence allegations in the document are not convictions.[^6]
Those qualifications do not destroy the story. They make its claim clearer.
A judgment does not remove a website from an access provider's resolver by magic. It changes legal conditions. It gives lawyers a new argument, prosecutors a reason to reconsider an instruction, and courts a precedent for decisions in other files. The evidence in the screenshot is a plausible snapshot of that institutional process.
It is not the spectacular moment in which the entire Swiss internet is instantly set free. It is more interesting: the quiet way a definition in a judgment begins to alter administrative conduct.
The two case numbers must never be collapsed into one. There is no reason to assume the two redacted domains belong to the same alleged fraud network as the twelve addresses in the original July 2025 order. There is no reason to assume that the November reversal was made by the same prosecutor or concerning the same accused people. The relationship visible in the document is a legal citation — a later case relying on a precedent — not an identity of factual circumstances.
For publication, obtain the original order or seek confirmation from the authority, preserve the redactions, and attach a provenance note. Until then, the chapter should use the cautious phrase *the posted document appears to show*. Readers deserve the difference between seeing a screenshot and verifying a signed original.
The visual evidence supports this reading of what the screenshot purports to show. It does not establish who first published the order, whether the image is complete, whether every named recipient received it, or whether corresponding DNS filters were subsequently removed. The identity of the accused person should not be inferred from the redactions. The underlying offence allegations in the document are not convictions.[^6]
A judgment does not itself remove a website from an access provider's resolver. It changes legal conditions. It gives lawyers a new argument, prosecutors a reason to reconsider an instruction, and courts a precedent for decisions in other files.
The two case numbers should not be collapsed into one. There is no reason to assume the two redacted domains belong to the same alleged fraud network as the twelve addresses in the original July 2025 order, or that the November reversal was made by the same prosecutor or concerning the same accused people. The relationship visible in the document is a legal citation: a later case relying on a precedent.
Until the original order is obtained or confirmed by the authority, the cautious phrase is *the posted document appears to show*. The redactions should be preserved.
<figure> <img src="/api/media/field-notes/website-that-was-never-seized-f07-en-372a16007b90a3fd63f6adb518a0e9957d3ea680837e9327cdfe374eef86f16d.webp" alt="Timeline from the August 2024 Vaud fraud investigation through the October 2026 Federal Supreme Court judgment, its November publication, and a separate 9 November document marked as unverified." /> <figcaption>Figure 6. Dated milestones. The timeline distinguishes the original fraud proceeding from the 9 November document in another file. The chart is schematic, not scaled to elapsed time.</figcaption> </figure>
The legal system is rarely cinematic at the moment it changes. Often the evidence is a sentence inserted into an ordinary administrative form.
A block had been ordered. A court said the general power did not exist. A later document, if genuine, shows another block being lifted.
The important fact is not that the machine once said no. It is that the reason for saying no could now be contested by law.
A block had been ordered. A court said the general power did not exist. A later document, if genuine, shows another block being lifted. The legal basis for such an order could now be contested.
## IX. The next law has already been proposed
The argument cannot end with the Supreme Court decision, because the court itself pointed toward Parliament.
In **May 2026**, months before the Supreme Court judgment, the federal government had opened consultation on a partial revision of Switzerland's Telecommunications Act. One part of the draft proposed a new **Article 6b**, aimed at combating fraud. Its wording matters more than the headlines about a general new ability to block scam websites.[^12][^13]
The court itself pointed toward Parliament. In **May 2026**, months before the Supreme Court judgment, the federal government had opened consultation on a partial revision of Switzerland's Telecommunications Act. One part of the draft proposed a new **Article 6b**, aimed at combating fraud.[^12][^13]
The draft says that, following information from **fedpol** and a sufficiently grounded suspicion of fraud under Article 146 of the Swiss Criminal Code, a telecommunications provider must block a **Swiss telephone number** for **no more than five days**. The next paragraph extends this model to the **registry operators** of **second-level domain names under domains for which Switzerland is responsible**. It also contemplates an order by the prosecutor or competent court to maintain a number or domain block until criminal proceedings end.[^12]
Read that slowly. The proposal speaks about phone-number providers and domain **registries**. It does not say that every Swiss access provider may be directed to manipulate DNS answers for any foreign-domain website. This distinction is particularly significant because online investment scams often use domains that are not under Swiss registry control. The draft therefore addresses part of the problem, but not necessarily the exact mechanism considered in the Supreme Court case.[^12][^13][^14]
Nor can the draft be treated as law already enacted. A consultation is a stage in legislation. The proposal may be amended, narrowed, expanded, rejected or enacted through subsequent processes. The source used here is dated **27 May 2026**. The commentary is limited to the state of the public material reviewed by **11 November 2026**.[^12][^13]
The harder policy question is what a lawful anti-fraud DNS-blocking system would require if lawmakers decided to create one.
It would need a clear definition of eligible offences and evidence thresholds. It would have to specify whether urgent action can precede judicial approval, how long it may last, when review must occur, and what proof is required to renew it. It would need to prevent a broad domain restriction when the alleged harm concerns a small path on a mixed-use service. It would need a route to notify affected site operators and providers without tipping off active fraudsters unnecessarily. It would need a functioning appeal, a way to reverse a block rapidly, and public information sufficient to discover accidental overblocking.
That is not a demand for absolute perfection before enforcement. A legal system routinely acts on provisional evidence. The point is that the powers must be bounded and reviewable in ways appropriate to their effects.
There is also a technical design choice. A domain registrar or registry, an access provider and a hosting provider have different points of leverage. A precise takedown by an operator that hosts a particular fraudulent page may have fewer collateral effects than a broad DNS intervention — but may be unavailable when the operator is abroad, unknown or uncooperative. A registrar-level action can affect the entire domain globally or across many resolvers, not merely the customers of a Swiss ISP. That may be stronger and also more consequential. Technical feasibility is not the same as proportionality.[^1][^13]
There is no single universal solution to online fraud. There are combinations of evidence preservation, investigation, bank intervention, warnings, hosting action, registry procedure, cross-border cooperation and — where expressly authorised and appropriate — carefully bounded access restrictions.
The mistake would be to treat every problem as a nail because a DNS block is one of the tools close at hand.
The proposal speaks about phone-number providers and domain **registries**. It does not say that every Swiss access provider may be directed to manipulate DNS answers for any foreign-domain website. That matters because online investment scams often use domains that are not under Swiss registry control. The draft therefore addresses part of the problem, but not necessarily the exact mechanism considered in the Supreme Court case.[^12][^13][^14]
The draft is not law already enacted. A consultation is a stage in legislation. The proposal may be amended, narrowed, expanded, rejected or enacted through subsequent processes. The source used here is dated **27 May 2026**. The commentary is limited to the state of the public material reviewed by **11 November 2026**.[^12][^13]
If lawmakers decided to create a lawful anti-fraud DNS-blocking system, it would need a clear definition of eligible offences and evidence thresholds; rules on whether urgent action can precede judicial approval, how long it may last, when review must occur, and what proof is required to renew it; a way to avoid a broad domain restriction when the alleged harm concerns a small path on a mixed-use service; a route to notify affected site operators and providers without tipping off active fraudsters unnecessarily; a functioning appeal; a way to reverse a block rapidly; and public information sufficient to discover accidental overblocking. A legal system routinely acts on provisional evidence. The powers would still need to be bounded and reviewable in ways appropriate to their effects.
A domain registrar or registry, an access provider and a hosting provider have different points of leverage. A precise takedown by an operator that hosts a particular fraudulent page may have fewer collateral effects than a broad DNS intervention, but may be unavailable when the operator is abroad, unknown or uncooperative. A registrar-level action can affect the entire domain globally or across many resolvers, including users who are not customers of a Swiss ISP. Technical feasibility and proportionality are separate tests.[^1][^13]
Online fraud is addressed through combinations of evidence preservation, investigation, bank intervention, warnings, hosting action, registry procedure, cross-border cooperation and — where expressly authorised and appropriate — carefully bounded access restrictions.
## X. The test of legitimate power
A useful government is not one that cannot act. A free society is not one in which every technical intervention requires unanimous consent. People are defrauded. Criminal investigations are urgent. Platforms can disappear and return under new names faster than ordinary legal procedure can move.
But urgency changes the *procedure* by which authority is exercised, not the basic need for authority.
The Supreme Court could have stretched a broad seizure concept over a new internet mechanism. It did not. Its refusal forced the uncomfortable but democratically appropriate question back into the open: if Switzerland wants a general power to order DNS blocking in criminal investigations, who should have it, for which offences, under which safeguards, and with what remedies?[^1]
The answer should not be smuggled into existence through technical defaults. Nor should it depend on whether the company receiving the order has an unusually determined chief executive or enough money to fight litigation. Rights are a poor feature of a network if they work only for users served by the one provider willing to appeal.
People are defrauded. Criminal investigations are urgent. Platforms can disappear and return under new names faster than ordinary legal procedure can move. Urgency may change the procedure by which authority is exercised. It does not remove the need for authority.
The Supreme Court could have stretched a broad seizure concept over a new internet mechanism. It did not. That left an open question: if Switzerland wants a general power to order DNS blocking in criminal investigations, who should have it, for which offences, under which safeguards, and with what remedies?[^1]
The answer should not depend on whether the company receiving the order has an unusually determined chief executive or enough money to fight litigation.
Here is the intervention test I would want lawmakers, regulators, operators and the public to be able to apply.
<figure> <img src="/api/media/field-notes/website-that-was-never-seized-f08-en-7dc05629a9eac397c68e5e5b0170e5abbcd45cc6c3ffa6ab040410a458a898d0.webp" alt="Checklist of seven proposed governance questions for a blocking law: purpose, evidence, authority, precision, duration, remedy and audit." /> <figcaption>Figure 7. Author's proposed seven-question checklist. These questions are policy recommendations; they are not a quotation of existing Swiss statutory conditions.</figcaption> </figure>
**Purpose:** Name the suspected harm and why access restriction is connected to stopping it. “Internet crime” is too broad to operate as a warrant.
**Evidence:** Identify what must be established, by whom, and at which level of confidence. An association by hyperlink is not necessarily evidence of complicity.
**Authority:** Specify the institution authorised to order the measure and the independent institution able to review it. Urgent provisional action must have a clock.
**Precision:** State the technical object — hostname, URL, domain, server, account or content — and forbid operators from silently widening the order merely because the requested precision is impossible.
**Duration:** Require expiry, renewal conditions, and the removal of restrictions when they are no longer justified.
**Remedy:** Give providers and affected operators a credible channel to contest incorrect or excessive measures, with a process for urgent correction.
**Audit:** Maintain reviewable records of orders, actual implementation, overblocking, reversals and the effect on third parties. Publish what can be made public without exposing victims or undermining investigations.
These are not arguments for doing nothing. They are arguments for building enforcement mechanisms that survive contact with both the law and the technology.
There is a temptation to describe all such issues through sovereignty alone. Sovereignty is a useful word, but it can conceal rival questions. Does sovereignty mean that Switzerland can compel firms inside its territory to alter internet access? That Parliament can define lawful powers instead of prosecutors improvising them? That Swiss residents can reach information even when an authority dislikes it? That fraud victims can expect meaningful protection? All of these claims point toward forms of legitimate self-government. None can simply cancel the others.
The central issue is not how much coercive power a nation possesses. It is whether the people subject to that power can trace it to law, challenge its use, understand its technical consequences, and hold its wielders accountable.
A state can be strong and still require a warrant.
A network can be open and still support lawful interventions.
A provider can obey the law and still demand that the law be clear.
Those are not contradictions. They are the operating conditions of a mature digital democracy.
Sovereignty can mean several things at once: that Switzerland can compel firms inside its territory to alter internet access; that Parliament can define lawful powers instead of prosecutors improvising them; that Swiss residents can reach information even when an authority dislikes it; that fraud victims can expect meaningful protection. None of those claims simply cancels the others. The people subject to the power need to be able to trace it to law, challenge its use, understand its technical consequences, and hold its wielders accountable.
## XI. What remains after the block
It is tempting to tell this story as a triumph of a small Swiss provider over the state. That version has a protagonist, a courtroom ending and an easy villain. It would also leave too much out.
The alleged fraud victims remain real people in the judicial account, even if their names are withheld and their losses have not been finally adjudicated in the material reviewed here. The state had a legitimate reason to look for preventative measures. Other legal systems and special Swiss statutes do permit certain forms of blocking. The technical instrument is not useless merely because it is bypassable. And a court victory on one legal question does not settle every related sanction, provider response, or pending case.[^1][^4]
The stronger ending is less flattering to everyone, and therefore more useful.
The alleged fraud victims remain real people in the judicial account, even if their names are withheld and their losses have not been finally adjudicated in the material reviewed here. The state had a legitimate reason to look for preventative measures. Other legal systems and special Swiss statutes do permit certain forms of blocking. The technical instrument is not useless merely because it is bypassable. A court victory on one legal question does not settle every related sanction, provider response, or pending case.[^1][^4]
The investigators were right to care about harm. The engineers were right that a hostname is not a URL path. The provider was entitled to ask whether general seizure rules could be extended to its resolver. The cantonal courts reached different intermediate positions. The Federal Supreme Court drew a line. The legislature may choose to create new powers, but it must make the choices visible.[^1][^5]
The website in the original case was never shown to have been seized by the DNS order. That was the point. It may have become harder to reach through certain networks, but the command did not transfer control of the website to the state. The public authority had tried to achieve the effects of a confiscation through a system that did not confiscate anything.[^1]
The distinction sounds almost philosophical until you imagine the next address. A fraudulent investment portal. A political website. A health forum. A newspaper. A whistleblower archive. A small business sharing an address with something illegal. Each has different facts. Each can nevertheless be touched by the same technical lever.
In October 2026, five judges in Lausanne did not decide that the state should ignore those differences. They insisted that the state must not erase them by renaming a block a seizure.
The site's servers were never the object of the order before the court.
The rule that authorised the order was the thing the judges took away.
The website in the original case was never shown to have been seized by the DNS order. It may have become harder to reach through certain networks, but the command did not transfer control of the website to the state. The public authority had tried to achieve the effects of a confiscation through a system that did not confiscate anything.[^1]
The next address might be a fraudulent investment portal, a political website, a health forum, a newspaper, a whistleblower archive, or a small business sharing an address with something illegal. Each has different facts. Each can nevertheless be touched by the same technical lever. In October 2026, five judges in Lausanne held that the state must not erase those differences by renaming a block a seizure.
---
## What this article does not establish
1. It does not establish guilt in any underlying fraud or arson case; investigative allegations remain allegations. 2. It does not identify the redacted domains in PE25.026767-JDZ or establish that the 9 November posted image is an authenticated original. 3. It does not establish a final legal outcome in the separate CHF 6,000 sanction reportedly imposed on Fredy Künzler. 4. It does not determine that Swisscom, Sunrise or Salt violated Article 12e FMG, nor does it claim that providers acted from improper motives. 5. It does not equate every Internet block with censorship; particular legally authorised measures must be evaluated individually. 6. It does not treat the May 2026 Article 6b consultation text as enacted law or as a general power over foreign domains. 7. It does not claim that switching resolvers always defeats every possible form of network restriction.
## Evidence and editorial method
This is Edition 2, published 12 November 2026. The facts, sources and figures are those of Edition 1, 11 November 2026; the prose has been tightened.
The evidence cut-off for this edition is **11 November 2026**. The controlling record is the published Federal Supreme Court judgment 7B_52/2026, supported by the published Valais decision P3 25 268 and the publicly available federal consultation draft. Contemporary reporting supplies identities, related litigation, the climate movement case, and responses from affected parties; these are attributed as reporting rather than disguised as findings of the Supreme Court. The alleged 9 November reversal is supported solely by the user-supplied image and is treated separately. Visuals are original explanatory graphics or concept art and never presented as documentary photographs.
The publication package includes a structured source register, claim-to-source map, visual briefs, alternative text, reproduction rights notes, social copy, and explicit outstanding checks. The SHA-256 generated for the draft in the package identifies the draft bytes only; it is not a certificate of factual correctness. The publication's official version and author sign-off should be recorded after editing.
## Sources
[^1]: Swiss Federal Supreme Court, *7B_52/2026*, judgment of 12 October 2026, published 4 November 2026; particularly facts B.a and reasoning 2.3–2.6. Official database access: [www.bger.ch](https://www.bger.ch/ext/eurospider/live/fr/php/aza/http/index.php?highlight_docid=2026-10-12-7B_52-2026&lang=fr). Readable mirror: [bgex.ch](https://bgex.ch/bger/7b_52-2026). [^2]: SRF, *Bundesgericht stoppt Netzsperren*, 4 November 2026. [www.srf.ch](https://www.srf.ch/news/schweiz/netzsperren-fehlende-gesetzesgrundlage-bundesgericht-stoppt-netzsperren). [^3]: Republik, *Netzsperren: Die Grossen kuschen, der Rebell kämpft*, 14 April 2026. Reporting on multiple orders, Switch's technical opinion, other providers and Grondements des Terres. [www.republik.ch](https://www.republik.ch/2026/04/14/netzsperren-die-grossen-kuschen-der-rebell-kaempft). [^4]: ICTjournal, *Init7 remporte un premier recours contre des blocages DNS*, updated 28 May 2026. [www.ictjournal.ch](https://www.ictjournal.ch/news/2026-05-28/init7-remporte-un-premier-recours-contre-des-blocages-dns-update). German parallel: [www.netzwoche.ch](https://www.netzwoche.ch/news/2026-05-29/update-init7-erzielt-gerichtlichen-erfolg-gegen-netzsperre). [^5]: Kantonsgericht Wallis, *P3 25 268*, order of 28 May 2026; original court text reproduced by Omnilex. [app.omnilex.ai](https://app.omnilex.ai/de/sources/court-decision/ch-vs-courts--p3-25-268/p3-25-268). Verify against the original court-hosted copy at the link labelled “Originalquelle öffnen”. [^6]: User-provided screenshot of a LinkedIn post containing an image purporting to show a Vaud prosecutor's decision, **PE25.026767-JDZ, dated 9 November 2026**, lifting seizures of two redacted URLs and citing 7B_52/2026. **Provenance: image only; no authenticated court or prosecutor original obtained.** Held only in internal evidence materials; not republished here. [^7]: Switch, *Kurzgutachten zu URL-Sperren*, signed opinion dated 23 February 2026; linked by Republik. [cdn.repub.ch](https://cdn.repub.ch/s3/republik-assets/repos/republik/article-strafverfolger-ordnen-netzsperren-an-die-grossen-kuschen-der-kleine-kaempft/files/f5502a21-fedb-4ab9-a78b-bca342ab6e9d/kurzgutachten_url_sperren-signed.pdf). Read-only reference; check reprint rights. [^8]: Inside IT, *Netzsperren: Init7 siegt vor Bundesgericht*, 4 November 2026. [www.inside-it.ch](https://www.inside-it.ch/netzsperren-init7-siegt-vor-bundesgericht-20261104). [^9]: Netzwoche, *Bundesgericht urteilt gegen Netzsperren ohne Sonderregelung*, 4 November 2026. [www.netzwoche.ch](https://www.netzwoche.ch/news/2026-11-04/bundesgericht-urteilt-gegen-netzsperren-ohne-sonderregelung). [^10]: SRF, *Was das Bundesgerichtsurteil zum Stopp von Netzsperren bedeutet*, 4 November 2026. Contains BAKOM and civil-society responses. [www.srf.ch](https://www.srf.ch/news/schweiz/keine-gesetzliche-grundlage-was-das-bundesgerichtsurteil-zum-stopp-von-netzsperren-bedeutet). [^11]: Swiss Telecommunications Act, Art. 12e, open internet, with statutory exceptions. Check the consolidated legislation and the exact applicable version on Fedlex before publishing any doctrinal interpretation. The Supreme Court also discusses specialised Telecom Act blocking arrangements in 7B_52/2026. Regulator commentary: [www.bakom.admin.ch](https://www.bakom.admin.ch/de/offenes-internet-netzneutralitaet). [^12]: Swiss Federal Office of Communications (BAKOM), draft amendment to the Telecommunications Act, **Vorentwurf vom 27. Mai 2026**, Art. 6b, p. 4. [www.bakom.admin.ch](https://www.bakom.admin.ch/dam/de/sd-web/97YpmKVP6XsG/Beilage%2001%20BG%20DE%20zu%20BRA%20UVEK.pdf). Consultation page: [www.bakom.admin.ch](https://www.bakom.admin.ch/fr/consultations-et-auditions). [^13]: BAKOM, explanatory report on the 2026 proposed Telecommunications Act amendments. [www.bakom.admin.ch](https://www.bakom.admin.ch/dam/fr/sd-web/xEvSxqGaHzDR/Beilage%2002%20Erl%C3%A4uternder%20Bericht%20FR%20zu%20BRA%20UVEK.pdf). [^14]: Legal commentary on the interaction of the draft Art. 6b and 7B_52/2026, BGer Update, 2026: [www.bger-update.ch](https://www.bger-update.ch/fr/?trk=public_post-text). Supports the narrower reading of the draft; the original draft text controls.