The VPN Bill Would Turn a Privacy Route Into a Blocking Point

You open a VPN because the local network should not get a clean list of the places you visit. Your traffic enters an encrypted tunnel, leaves through a server run by the VPN company, and continues toward the site you chose. That route gives the VPN provider a powerful position. A bill now before the US House would ask some providers to use that position as a checkpoint.

Representative Darrell Issa introduced H.R. 10364, the American Copyright Protection Act of 2026, on 14 September 2026. The bill would create a court process for declaring certain foreign online locations to be piracy sites. A copyright owner could then ask the court to order named broadband providers, domain-name resolvers, and VPN services to take “commercially reasonable steps” that stop people in the United States from reaching the site (GovInfo: H.R. 10364, introduced text).

The proposal remains a bill. As of 5 October 2026, it has been introduced and referred to the House Judiciary Committee; it has not become law (GovTrack: H.R. 10364 status). That distinction matters because no VPN has acquired a new federal blocking duty merely because this story exists.

The useful question sits one layer deeper. What changes when a privacy route also becomes an enforcement point? The answer affects more than copyright. It reaches the architecture of VPNs, the damage caused by a wrong block, and the promises you should expect a provider to make before a court order arrives.

What H.R. 10364 would actually create

The bill uses two court stages. First, a copyright owner could bring an action seeking a declaration that a foreign online location is a “foreign piracy site.” The location must be public, operated outside the United States or not reasonably located inside it, and identifiable through network addresses, domain names, or similar identifiers. The copyright owner would have to prove the statutory conditions by a preponderance of the evidence (GovInfo: H.R. 10364, sections 1601 and 1602).

Those conditions include infringement that harms or will harm the copyright owner, efforts to notify the site operator and registrar, and one of three descriptions of the service. The site may be primarily designed for infringement, marketed for infringement, or have “only limited commercially significant purpose or use” beyond giving access to infringing material. That final phrase is wider than a test limited to a site with no significant lawful use.

The accused site may never appear at the first stage. Its operator may be abroad, hidden, unreachable, or unwilling to submit to a US court. The judge may appoint a special master to test the claimant’s evidence, but the bill says “may,” not “shall.” The Electronic Frontier Foundation’s 2 October analysis identifies this as the central process problem: a court could attach the piracy label before anyone representing the site or its users argues the other side (EFF: Congress Has Another Site-Blocking Bill, And This One Targets VPNs).

The second stage names the intermediaries. Once the declaration exists, the copyright owner could petition for a blocking order against the providers it believes are needed to reduce US access by a commercially significant amount. Each named provider would receive the petition and could respond. Under the ordinary timetable, the court could not issue the order until at least 14 days after filing.

A resulting order would require each named provider to take all commercially reasonable steps to prevent US users from reaching the declared site. Implementation would ordinarily be due 14 to 30 days after notice. An order could last up to 12 months, with a process for extensions. The bill leaves the technical method to each provider rather than prescribing a particular filter.

The definition of provider is where this story meets privacy tools. It expressly includes broadband internet access, domain-name resolution, and virtual private networks. Providers with fewer than 100,000 monthly US users or subscribers are excluded, as are premises such as libraries, universities, restaurants, shops, and airports that buy connectivity for people on site. Ars Technica reported the same scope after reviewing the circulated bill text in September (Ars Technica: Republican bill would order ISPs, DNS providers, and VPNs to block piracy sites).

This is a court-order system, not a private notice that instantly switches off a domain. The difference is real. So are the safeguards: provider notice, a chance to object, judicial consideration of technical feasibility and incidental harm, a public registry of active orders, bonds posted by claimants, correction procedures, and compensation in some wrongful-blocking cases. A serious reading should acknowledge all of them.

Those protections reduce risk after a claim begins. They do not remove the architectural choice. The United States would be creating a process whose intended output is blocked access, and large VPN services would sit among the entities expected to produce it.

A VPN becomes useful because it occupies the middle

A VPN is often sold as a cloak. The mechanism is less romantic. Your device makes an encrypted connection to a provider’s server. The coffee-shop network or home internet provider can usually see the VPN connection, while the destination sees a connection arriving from the VPN server. The VPN company handles the traffic at the seam between those two legs.

That seam changes who can observe and control the route. Your access provider loses the simple destination view it once had. The website loses your household network address. The VPN gains a position from which it can see destination information, apply routing rules, reject traffic, or keep records, depending on its design and the protocols in use.

The provider does not need to break the tunnel to block a destination. It terminates the tunnel. From there, it can refuse to route traffic toward an address, interfere with a name lookup it operates, or apply another control chosen for its network. The exact method matters because a domain, an address, and a web page are different things.

Suppose an order identifies example.test by domain name. A provider might block its own resolver from returning an address. A customer using a different resolver inside the tunnel may still resolve it. The provider could then block the destination addresses, but one address may host many unrelated sites, while the target may move across addresses or use a large content-delivery network. Filtering a specific encrypted web path becomes harder because modern HTTPS hides the page content and often much of the request from the network.

“Commercially reasonable” does heavy work here. The bill tells the provider to choose the method and tells the court to consider burden, feasibility, efficacy, incidental harm, and public benefit. It does not make those factors agree. A narrow method may be easy to bypass. A broad method may catch lawful services sharing the same infrastructure.

Blocking also collides with the reason some people choose a VPN. A source may use one to keep a hotel or access provider from learning which newsroom they contacted. An activist may need stable access to a site that is controversial in one jurisdiction and lawful in another. A traveller may use a familiar exit country to avoid hostile local filtering. None of these uses makes every destination lawful, but each raises the cost of a mistaken or opaque block.

The practical lesson is blunt. A VPN provider already occupies a trust boundary. H.R. 10364 would make one possible exercise of that trust explicit: a sufficiently large provider could be named in a US order and required to stop its US users from taking a specified route.

The first mistake can spread far beyond the target

Site blocking looks precise in a legal caption. Networks are shared, layered, and constantly changing. The court names an online location; the provider must translate that label into rules that machines can enforce. Every translation chooses a unit of collateral damage.

A domain-name block can affect every page and service beneath the name. An address block can affect unrelated customers on shared hosting. A content-delivery network may place thousands of sites behind a changing pool of addresses. A target can move, while an abandoned address is reassigned to an innocent service. Cached rules survive unless someone owns their removal.

The bill anticipates some of this. Copyright owners would have to monitor the continuing accuracy of domains, addresses, and URLs that supported the order, then notify the court and providers of material changes. Providers could temporarily suspend implementation to investigate accidental blocking. An unaffiliated third-party service that was wrongly restricted could ask the court to modify the order and, in some cases, seek compensation capped at $250,000 if the copyright owner caused the error (GovInfo: H.R. 10364, sections 1604 to 1606).

That is better than pretending errors will never occur. The repair path still starts after access has broken. A small forum, health service, campaign archive, or independent publisher may first discover the order when its US audience disappears. Finding the responsible provider, locating the case, hiring counsel, proving injury, and waiting through a response period all consume time.

Experience abroad supplies the warning. EFF’s October analysis cites Italy’s Piracy Shield blocking hundreds of benign sites, including a convent and a telehealth platform, and a Spanish blocking programme that swept up domains belonging to organisations such as Greenpeace and Harvard University (EFF: Congress Has Another Site-Blocking Bill, And This One Targets VPNs). Ars reported research finding at least hundreds of legitimate sites affected by Italy’s system and described outages involving payment, healthcare, and small-business infrastructure in European blocking disputes (Ars Technica: Republican bill would order ISPs, DNS providers, and VPNs to block piracy sites).

Those examples do not prove that every US order would fail in the same way. They show that collateral damage is an observed property of blocking systems, not a theoretical objection invented to end debate. Shared infrastructure turns a mistaken identifier into other people’s outage.

Speed makes the trade sharper. H.R. 10364 defines a time-sensitive event as a work distributed or performed at the same time as, or within 24 hours after, its first authorised US release. A court could modify deadlines for good cause, including these events, while remaining consistent with due process. Live sport is the obvious case: a block that arrives after the match has little commercial value.

Fast relief gives everyone less time to test identifiers and predict shared-hosting effects. A provider may prefer a broad rule that clearly satisfies the order over a delicate rule that misses some traffic. The bill asks courts to weigh collateral harm, but the technical facts can change between the hearing and deployment.

A blocklist therefore needs the discipline of production security work. Each entry needs a source, owner, scope, activation time, expiry, test, rollback path, and record of what it actually affected. A PDF order by itself cannot provide that operational receipt.

Due process cannot be bolted on after the route disappears

The strongest defence of the bill is procedural. Copyright owners would go to court. Providers would get notice. Judges would weigh feasibility and harm. Claimants would post bonds. Site operators could later seek rescission. A public registry would show active orders. These features deserve credit because they create places to challenge a bad result.

The weak point arrives before the provider dispute. The first declaration asks whether the foreign location qualifies as a piracy site. The copyright owner presents the case. The operator may be absent. A special master is optional, and even an appointed expert does not represent the missing site or the people who lawfully use it.

That asymmetry matters because the later provider hearing starts with the declaration already made. A VPN can argue that including its network would be burdensome, ineffective, or harmful. It is not necessarily standing in for every lawful publisher, researcher, or user attached to the accused location. Its incentives centre on its own network, costs, and liability.

The operator’s route back carries a price. Under one rescission path, it submits to the court’s jurisdiction, waives several objections, posts a bond sufficient to support a later judgment, and complies with other orders to stop the alleged activity. Another path lets it show that the earlier findings lack evidentiary support. Either way, an overseas operator that ignored the opening case may need US counsel and money after the label begins doing damage.

The third-party correction process is also reactive. An innocent service sharing infrastructure can move to restore access only after the order restricts it and causes injury. The court must consider less disruptive changes that preserve the wider order. Compensation is possible when the claimant caused the error, but money later does not restore a confidential conversation, a deadline, or trust lost during an unexplained outage.

A public registry helps people investigate. It works only if users can distinguish an ordered block from a technical fault. The bill allows providers to tell users and the public about an order, but it does not require them to do so. A failed connection could look like downtime, censorship by a local network, a provider fault, or a legal block. Ambiguity favours the party controlling the route.

The right transparency rule should be mandatory and specific. A provider should return a clear notice where the protocol allows it, name the court and case, link to the active order, state the identifier it blocked, record the time, and offer a route for reporting collateral damage. Sensitive users should not have to file a support ticket that exposes more account information merely to discover why a site vanished.

Transparency does not cure a bad order. It shortens the interval during which nobody knows who made the decision. That interval is where quiet censorship thrives.

Privacy depends on exits, not slogans

People often judge a VPN by encryption, speed, and a promise about logs. Site blocking adds another question: what happens when the provider is told to close a route? The answer should live in a policy written before the first order arrives.

A useful policy would state which legal entity operates each service, where orders are received, how the provider verifies scope, and whether it challenges orders that conflict with its architecture or affect lawful traffic. It would explain how users are notified, how errors are reported, and how stale rules are removed. A transparency report should count orders received, challenged, narrowed, implemented, and later corrected.

Server labels alone do not settle jurisdiction. A provider may advertise an exit in Iceland while the customer contracts with a US entity, the control plane runs elsewhere, and staff deploy network rules globally. The country displayed beside a connect button describes one machine’s apparent location. The company’s legal and operational layout describes who can change it.

The distinction becomes important under H.R. 10364 because the order concerns access from the United States. A provider must decide which sessions count as US access. Account country, payment record, device location, entry address, and exit server can disagree. The bill does not hand users a simple answer about a US subscriber travelling abroad or a foreign subscriber connecting through a US entry point.

A provider could avoid collecting extra location data by applying a rule more broadly. That protects against the compliance miss and enlarges the censorship footprint. It could build finer location controls, which narrows the block and creates another collection system. Enforcement pressure often converts uncertainty into surveillance.

This tension deserves an answer before enactment. The bill directs a later Copyright Office report two and five years after the law takes effect, but users need safeguards at deployment time. Data minimisation, clear geographic scope, short retention, independent testing, and public error counts should be conditions of a blocking system rather than lessons recovered from its first failure.

You also need an honest threat model. A commercial VPN can protect your traffic from a hostile hotel network and hide destination addresses from your ordinary access provider. It cannot promise access against every order served on the VPN company. It cannot protect an account that identifies you, a browser that carries old cookies, or an endpoint already under control.

The Anonymity Playbook makes the same point from another direction: privacy comes from separating trust and knowing where each tool stops. A route through one company is still one company’s route. The hard question is how quickly you can move when that route no longer serves your purpose.

That does not require a pile of five VPN subscriptions. It requires a tested alternative appropriate to the stakes. For ordinary browsing, waiting or using another lawful connection may be enough. A journalist protecting access to a source archive needs independent channels, local copies of permitted material, and a plan that does not announce the relationship through frantic troubleshooting after a block begins.

What to check before this becomes an emergency

A bill at committee stage does not justify panic or a weekend spent rebuilding every device. It does justify asking providers and lawmakers questions while the design can still change. The right preparation leaves a receipt instead of a new superstition.

1. Confirm the legal status. As of 5 October 2026, H.R. 10364 is introduced legislation, not an active blocking law. Read the current bill page before acting on a later headline because amendments can change covered providers, procedure, and remedies (GovTrack: H.R. 10364 status).

2. Read your VPN provider’s legal and transparency material. Look for counts of blocking orders, jurisdiction, notification practice, challenge criteria, and correction history. A page that says only “we comply with lawful orders” leaves the operational questions unanswered.

3. Ask how a block is shown. The provider should tell users when a court order caused the failure, identify the order, and offer a reporting path. Silent packet loss is cheap for the provider and expensive for everyone diagnosing it.

4. Ask how scope is determined. A rule aimed at US access needs a documented way to decide which traffic falls inside it. The answer should avoid turning every VPN session into a new location-tracking event.

5. Test one lawful fallback. Keep the test boring. Confirm that a second route reaches a public, non-sensitive service, then record what changed in DNS and visible network address. Do not wait for an urgent investigation or source conversation to discover that both routes depend on the same company.

6. Keep critical lawful material available where permitted. Organisations that depend on a public archive, documentation site, or publication platform should know what can be mirrored or retained under licence. Availability planning is easier before a disputed domain joins a blocklist.

7. Report collateral damage with evidence. Save the time, provider, resolver, exit region, affected name, returned error, and a harmless test from another network. Avoid sending private browsing history. A compact record helps the provider distinguish an order from a routing fault and gives an innocent operator something usable in court.

8. Press for a narrower law. At minimum, the process should require meaningful representation at the declaration stage, mandatory user notice, public machine-readable orders, precise expiry, independent measurement of collateral damage, and a fast remedy that does not make innocent services finance their own restoration.

These steps address different risks. A fallback preserves access. Transparency assigns responsibility. Scope limits reduce collection. A correction record helps repair a wrong block. None makes a VPN invisible to law or immune from infrastructure mistakes.

The people with higher stakes should add one more exercise. Draw the route from device to access network, VPN entry, VPN control plane, exit, resolver, hosting provider, and destination. Mark the organisations that can observe or interrupt each edge. If the same company or jurisdiction controls several edges, the diagram will expose the concentration that a product label hides.

The real policy choice is where failure lands

Copyright owners have a genuine problem with foreign services that ignore ordinary notices and court claims. Live streams can move faster than conventional litigation. H.R. 10364 tries to answer that problem by reaching intermediaries that are visible, domestic, and able to interrupt traffic.

That choice moves the enforcement error. Instead of failing to stop some infringing visits, the system can fail by stopping lawful visits, collecting more location data, or making privacy providers less predictable. Bonds and correction motions decide who may recover after the mistake. They do not decide whether the mistake happens during a confidential deadline.

The bill contains more process than a caricature of instant private censorship. Its court stages, provider objections, harm factors, public registry, expiry, and correction mechanisms are material. The missing adversary at the first declaration and the inclusion of VPNs remain material too. Both facts can fit in the same sentence without cancelling each other.

A VPN is a route through a trusted intermediary. Any law that turns that intermediary into a blocking point should demand narrow identifiers, visible decisions, measured collateral damage, and quick rollback. Otherwise, “commercially reasonable” will be judged after each provider has already chosen how much of the network to close.

You cannot buy your way out of every legal boundary with an app. You can choose providers that publish their boundaries, keep an independent route for work that matters, and refuse the comforting fiction that encrypted transport guarantees reachability. Privacy survives through explicit limits.

For one practical security and privacy note each month, the newsletter signup is here on the site. One email per month.

Sources

• GovInfo: H.R. 10364, American Copyright Protection Act of 2026, introduced text, accessed 2026-10-05
• GovTrack: H.R. 10364, status and legislative history, accessed 2026-10-05
• Electronic Frontier Foundation: Congress Has Another Site-Blocking Bill, And This One Targets VPNs, accessed 2026-10-05
• Ars Technica: Republican bill would order ISPs, DNS providers, and VPNs to block piracy sites, accessed 2026-10-05