Congressman Michael Rulli of Ohio recently introduced a federal GRANITE Act (Guaranteeing Rights Against Novel International Tyranny and Extortion) to respond to what is commonly referred to as the foreign censorship threat. The legislation employs the approach devised by lawyer Preston Byrne to fend off, or even punish, what Byrne calls foreign “censorship demands.” But the problem is that foreign censorship is no longer just a threat. As a result of the European Union’s Digital Services Act, it is a widespread reality across all major online platforms. And the GRANITE model cannot stop it.

This model may prove serviceable in rebuffing the extraterritorial demands that British authorities might make under the Online Safety Act, which inspired Byrne’s idea, as his account of the GRANITE model’s genesis makes clear. But the Online Safety Act—ostensibly aimed at protecting minors and requiring the suppression of child sexual abuse material and terrorist content—has not been a major driver of online censorship.

After all, though British authorities may be acting outside their jurisdiction in attempting to enforce the act, the content targeted for suppression is, as a rule, also illegal in the United States and most of the Western world. Other content—for instance, legal pornography—is not supposed to be censored but merely made inaccessible to minors under the law. This is what is reportedly at issue in the famous 4chan case that gave rise to the GRANITE model.

For both substantive and economic reasons, the Digital Services Act (DSA) is a far greater driver of online censorship than the Online Safety Act—in fact, it is virtually the all-encompassing driver of online censorship.

To start with, the DSA’s standard of illegality is based on European laws, which cover a far wider swathe of discourse than American or British law. Under German law, this includes not just allegedly racist or xenophobic “hate speech” but even mere “insults.” More importantly still, the DSA’s censorship provisions target even legal speech if it is deemed in any way socially “harmful,” chiefly alleged “disinformation.” This is what has made the DSA a veritable turbocharger of online censorship.

Moreover, and this is the economic aspect alluded to above, the E.U. market is roughly eight times larger than the U.K. market, giving the E.U. far greater economic and hence political clout. If the requirements of British authorities become too onerous, major platforms might just leave the British Isles. But they will accept a far higher level of pain to remain on the E.U. market.

Regrettably, a GRANITE Act will have no impact on DSA-driven online censorship.

The most fundamental reason is that the DSA does not, as a rule, involve discrete censorship demands from governments. Under Article 9 of the DSA, E.U. member state governments may issue orders to platforms to take action on illegal content. But these represent an infinitesimally small part of the total censorship taking place under the DSA. For instance, per X’s latest so-called “DSA Transparency Report,” in the first half of this year X received a grand total of just 12 such government orders, involving only 37 items of content. This represents less than 0.01% of the total censorship taking place on X under the DSA.

How do we know? Well, because we also have data on the censorship occurring via the so-called notice-and-action mechanism (Article 16): the heart and soul of the DSA. The notices are not submitted by governments but rather by literally any “individual or entity” in the E.U., and submitters are not even required to disclose their identities. Although government-certified “trusted flaggers” also use the mechanism, the overwhelming majority of notices—over 99%—come from the unknown “individuals and entities.” The platforms themselves evaluate the flagged items in terms of their potential illegality and/or “harmfulness” and take action accordingly.

For instance, according to X’s data for the first half of 2026, it received a total of 492,493 notices comprising 755,395 items of content. The platform took action on 147,370 items determined to be “illegal”—that is, per some European law or another—and also on no less than 319,157 items that were not determined to be illegal. Whereas illegal items are supposed to be removed outright or at least blocked in the relevant jurisdiction, “legal but harmful” items receive the supposedly more delicate treatment of having their visibility restricted.

Overall, 466,527 items were censored in response to notices as compared with 37 censored in response to government orders.

Digging down a bit further into the data, especially into the 319,157 items that were censored despite not being illegal, we find that 82,880 items were classified as allegedly “harmful speech,” of which 39,232 were classified as “hate speech,” although apparently not illegal hate speech. Another 4,587 items constituted alleged “misinformation” affecting “civic processes and elections.” Another 630 items were classified as representing a “risk to public health,” and 86 were tagged as an environmental risk. These too presumably consist of alleged misinformation or disinformation.

This data, moreover, is incomplete. It does not include alleged misinformation or disinformation that the platform has censored proactively to “mitigate systemic risks,” as required under Articles 34 and 35 of the DSA. Nor does it include account-level actions, which the platform acknowledges undertaking and which affect every post on a given account.

Hence, the real number of items being censored could go well into the millions. From the censors’ point of view, the great beauty of this censorship is that most of it takes place in the form of visibility filtering, which is itself largely invisible.

How is a GRANITE Act going to stop all that? The problem is that the DSA does not give E.U. member states or the European Commission unlimited power to demand censorship of specific content. Rather, it creates a general legal expectation that the platforms themselves should censor. This has come to be known as “content moderation” in the terms of the trade. The DSA gives the Commission—and, in an ancillary fashion, also E.U. member states—oversight authority to determine whether the platforms are censoring “correctly” and censoring enough.

The role of the public authorities is, so to speak, ex post, not ex ante. The platforms are not told what to do. They have to intuit what the oversight authorities want censored. Hence, to ensure that they are reliably aligned, the platforms are in constant contact with European authorities, either collectively in the context of the “Codes of Conduct” on hate speech and disinformation or individually, tête-à-tête.

Even X, which Elon Musk theatrically withdrew from the E.U.’s Disinformation Code, has acknowledged in regulatory submissions that it “proactively engage[s] and exchange[s] information with the European Commission, the European External Action Service (EEAS), the European Parliament, and Member States’ key authorities” in its efforts to “mitigate systemic risks.”

Indeed, as I recently reported, platform operators and Google have also been meeting regularly with German authorities, including the German DSA oversight authority, on these same subjects. How does the American government allow this? President Trump’s Executive Order 14149 specifically prohibits American authorities from discussing disinformation with the platforms. How are American companies permitted to engage in the same sort of discussions with a foreign power?

The GRANITE Act approach attempts to create new protections for American platforms in American courts (“the shield”) and even a new right (“the sword”: the right to sue foreign governments). This may help smaller, non-commercial platforms that do not care about exploiting the European market and can simply withdraw from it.

But the DSA makes the major social media platforms complicit in European censorship. They are not the victims—they are accomplices. They do not need protection—they need punishment. The victims are ordinary Americans who can no longer express themselves freely online and have thus been robbed of their First Amendment rights.

If the E.U. can punish the platforms for not censoring, then the U.S. can punish them for censoring—and the U.S. is a big enough market that it can make them hurt just as much. The platforms need to be forbidden from complying with the DSA: no consultations with the European Commission or E.U. member state governments on hate speech or disinformation and no removal or suppression of any First-Amendment-protected speech in response to DSA notices or to satisfy any other requirements of the law. Far from being an acceptable way to reconcile free-speech principles with European requirements, visibility filtering is just a more insidious form of censorship and should be outlawed altogether.

I have laid out some elements of the required counter-legislation—an American Anti-Digital Services Act—in the final section of my Claremont Review of Books Spring 2025 essay. As I put it there, either the E.U. should be made to back down, or the platforms should be made to choose which market they want to be part of. If they know what is good for them—namely, a free market, including what we used to know as the free market of ideas—they will choose the U.S. If they are not forced to make the choice, they will take the path of least resistance and continue to censor Americans to satisfy a foreign power.