The OIC Blasphemy Campaign: The International Speech Ratchet

On September 30, 2005, the Danish newspaper Jyllands-Posten published twelve editorial cartoons depicting the Prophet Muhammad. The stated purpose was to test self-censorship in the Danish press. The test was conclusive. Roughly 200 people died in the protests, riots, and reprisal attacks that followed over the next year. Embassies burned. Cartoonists went into hiding. Yale University Press, in 2009, published an academic book about the controversy and removed every image of the cartoons from the text without informing the author.

The Organization of Islamic Cooperation had been working the same problem at the UN for six years before the cartoons were drawn. The cartoons accelerated the campaign. Then the campaign failed. Then it rebranded. And the architecture it built, the international standard for restricting “harmful” content, survives in the upload filters of every major content platform.

The Defamation of Religions Resolutions

From 1999 through 2010, the OIC pushed an annual resolution at the United Nations Human Rights Council and the General Assembly under the heading “Combating Defamation of Religions.” Pakistan introduced it on the OIC’s behalf. The substantive claim was that criticism of religion (initially of all religions, in practice almost entirely Islam) should be classified as a human rights violation.

The resolutions were non-binding. They passed every year. The margin declined every year. The General Assembly vote went from 101-53 in 2005 to 79-67 in 2010. Western democracies consistently voted against. The Library of Congress’s tracking of UN blasphemy resolutions documents the decline year by year. By 2010, the OIC could read the trend line. The campaign was losing ground every cycle.

In 2011, the OIC abandoned the framing and switched to a new vehicle.

Resolution 16/18

Resolution 16/18, adopted unanimously by the UN Human Rights Council in March 2011, was the rebrand. The text shifted from protecting “religions” against defamation, a category the First Amendment treats as incoherent because religions are not persons and cannot be defamed, to protecting “persons based on religion” against incitement to discrimination, hostility, or violence. The framing now mapped onto language the United States, the European Union, and most Western democracies could accept.

Secretary of State Hillary Clinton and OIC Secretary General Ekmeleddin İhsanoğlu jointly launched the Istanbul Process in July 2011 as the implementation mechanism. The Istanbul Process meetings have continued more or less annually, hosted by different governments, producing best-practice recommendations on training law enforcement, monitoring incitement, and developing what the resolutions call “interfaith dialogue.”

The Rebranding Debate

Two readings of Resolution 16/18 have coexisted since 2011.

The Brookings Institution reading: the text is carefully limited to incitement to imminent violence, which already meets the Brandenburg v. Ohio standard for unprotected speech under US First Amendment doctrine. Resolution 16/18 is, on this reading, a consensus document that codifies existing practice rather than expanding it.

The Jonathan Turley reading: the resolution creates institutional pressure to restrict speech across a much broader category than imminent violence, regardless of the formal text. Turley’s October 2012 congressional testimony argued that the implementation process, the recurring Istanbul Process meetings, and the diplomatic pressure on Western governments to demonstrate compliance produce a chilling effect on speech that does not appear in the words on the page.

Pakistan’s ambassador, in remarks reported in the implementation period, said the OIC would “test” the 16/18 framework and return to binding mechanisms if it proved insufficient. Read against the original Defamation of Religions resolutions, that statement is not subtle.

There is no documented direct pipeline from Resolution 16/18 to specific Western legislation. There are convergent interests. The European hate-speech traditions that predate the OIC campaign moved in a direction the OIC favored. The content-moderation policies of every major platform adopted “attacks based on religious affiliation” categories that overlap with the resolution’s framing. Convergence is not causation. It is also not unrelated.

The Blasphemy Ratchet in Practice

The Danish cartoon crisis (2005-2006) produced roughly 200 deaths and the Yale University Press self-censorship. The pattern was set: violence over depictions of the Prophet, followed by institutional self-censorship in the West, followed by a new baseline for what gets published.

The Charlie Hebdo attack on January 7, 2015, killed twelve people at the satirical magazine’s Paris offices: journalists, cartoonists, a maintenance worker, a police officer. France’s response was the Intelligence Act passed in July 2015. Widely called the “French PATRIOT Act.” It authorized bulk metadata collection, real-time internet surveillance, and the installation of “black boxes” on ISP networks to detect suspicious patterns using algorithmic analysis. No individualized warrant required. The act had been drafted before Charlie Hebdo and stalled in parliament. After the shooting, it passed with minimal opposition. The New York Times, the BBC, CNN, and the AP chose not to show the cartoons in their coverage. Six PEN members declined to attend the organization’s gala in protest of the Free Expression award given to Charlie Hebdo.

The Rushdie fatwa, issued in 1989 over The Satanic Verses, was reaffirmed and the bounty increased to $600,000 in 2016. Salman Rushdie was stabbed fifteen times in August 2022 at the Chautauqua Institution in upstate New York. The attacker was twenty-four years old and had been born more than a decade after the fatwa was issued.

Pakistan’s blasphemy laws (Penal Code sections 295-B and 295-C, with the death penalty added in 1986) have been used to charge thousands. More than 80 people have been murdered by mobs over blasphemy allegations since 1990 according to the International Commission of Jurists tracking. Asia Bibi, a Christian farm worker, spent eight years on death row before being acquitted in 2018 and was unable to leave Pakistan safely until 2019. Punjab governor Salmaan Taseer was assassinated by his own bodyguard in January 2011 for publicly supporting Bibi’s appeal. The bodyguard was hailed at his trial by lawyers throwing rose petals.

The Obama Pivot

In September 2012, President Obama addressed the UN General Assembly two weeks after the attack on the US consulate in Benghazi and during the protests over the Innocence of Muslims video. The line that the Istanbul Process advocates noticed was specific: “The future must not belong to those who slander the prophet of Islam.”

The sentence was, in its immediate context, an attempt to manage a foreign policy crisis. It was also, in its theological framing (the prophet, not a prophet, with the definite article) the President of the United States adopting the OIC’s preferred linguistic register at the UN podium. The line did not appear in subsequent State Department messaging. It did not need to. It had already been said.

The Content Moderation Echo

Facebook’s community standards, in their multiple revisions through the 2010s, developed explicit categories for “attacks based on religious affiliation.” The training materials for content moderators in the Sama and Majorel contracts in Kenya and Ghana, documented in TIME’s reporting on the moderators’ working conditions, included examples of religion-based content that triggered removal under platform policy.

Each violent incident produced a new layer of self-censorship. The platforms tightened. The new baseline did not loosen when the news cycle ended. This is the same mechanism as the Yale University Press redaction of the cartoon images: the institution decides, in advance of any specific demand, to remove the content that might trigger a response. The decision is presented as editorial judgment. The editorial judgment converges on the OIC’s preferred output.

The Terrorism-to-Content-Infrastructure Pipeline

ISIS media in 2014-2017 was the pretext for the construction of the Global Internet Forum to Counter Terrorism (GIFCT), launched in 2017 by Facebook, Microsoft, Twitter, and YouTube. The core technology was a shared hash database that allowed any participating platform to immediately remove content that had already been removed by another platform. The original scope was ISIS beheading videos. The mechanism, automated removal of content matched by perceptual hash, was content-agnostic. The database could match anything anyone added to it.

The Christchurch Call followed in May 2019, launched by Jacinda Ardern and Emmanuel Macron after the New Zealand mosque shootings that were livestreamed on Facebook. The shooting video had been viewed roughly 4,000 times before takedown. Fifty-six countries plus the European Commission and the major tech companies endorsed the Call. The United States declined to sign under the Trump administration, citing First Amendment concerns. The Biden administration endorsed it in 2021.

The EU Terrorist Content Regulation (Regulation 2021/784) followed in 2021, with a one-hour removal requirement and fines up to four percent of global turnover. Platforms must implement proactive detection, which in practice means upload filters that scan content before it is published.

Germany’s NetzDG, passed in 2017, modeled its takedown mechanism on counterterrorism removal and extended the scope to twenty-two content categories. The structure was copied by Russia, Singapore, Australia, Turkey, Venezuela, Honduras, and more than twenty other governments. The Russian version, in particular, retained the architecture and dropped the speech-protective limitations.

Once the upload filter exists, it can filter anything. That is the durable point. The terrorism justification built the infrastructure. The infrastructure is content-agnostic. The next operator decides what counts as harmful.

The EU DSA Completes the Layer

The EU Digital Services Act, effective February 2024, transforms platform content moderation from a voluntary practice into a legal obligation with fines up to six percent of global turnover. Very Large Online Platforms must conduct systemic risk assessments for disinformation, election integrity, public health, and the protection of minors. The Article 22 trusted-flagger system designates organizations, appointed by national authorities, whose content reports receive priority treatment from platforms. By mid-2025, 34 trusted flaggers had been appointed across the EU. In Germany, the Respect reporting office receives roughly 95 percent of its funding from the German federal government’s Demokratie Leben programme. A government-funded NGO, designated by the government as a trusted flagger, whose reports receive priority treatment from platforms.

The Brussels Effect does the rest. Platforms that must comply with EU law for EU users implement the same policies globally. The EU’s definition of “illegal content” (which varies by member state and includes “hate speech,” a category with no First Amendment equivalent) becomes the de facto global standard without any non-EU citizen having voted on it.

The OIC could not get the UN to make blasphemy a human rights violation. The EU has produced a content-moderation regime that, in operational practice, removes content the OIC would have flagged under the original defamation-of-religions framing.

The campaign at the UN failed. The cartoons stayed in the original publications. The First Amendment held. The Western democracies that voted no kept voting no. The 2010 resolution was the last under the defamation framing.

The architecture moved. Resolution 16/18. The Istanbul Process. The GIFCT hash database. The Christchurch Call. The EU TCR’s one-hour removal mandate. NetzDG and its twenty-five-country copy list. The DSA’s trusted-flagger system. Each layer was built by a different process with a different stated justification. Each layer outputs the same operational reality: faster, broader, less reviewable removal of content that any of the layered authorities considers harmful.

The Yale University Press redacted the cartoons before any demand was made. That is what the ratchet looks like at the end. Not enforcement. Anticipation.

Click.

The receipts (free, on this site): AI moderation · deplatforming · Three Legal Traditions and the Control Grid — why the First Amendment held while France legislated and the EU built the DSA: the constitutional layer under the blasphemy-axis divergence

This research appears in The Ratchet, Chapters 8 and 14.