The Infrastructure: Who Decides What Stays Up

Murthy v. Missouri, and EFF/FIRE counterpoints are all here. –>

The question of who decides what stays up has a second, less visible layer: the network of organizations that tell platforms what should come down.

Between 2019 and 2024, a constellation of think tanks, government-adjacent organizations, and advocacy groups coalesced into what their critics call the censorship industrial complex and their defenders call the disinformation research ecosystem. The institutions include the Stanford Internet Observatory, the Atlantic Council’s Digital Forensic Research Lab, the Election Integrity Partnership, the Global Disinformation Index, the Global Engagement Center, NewsGuard, the Center for Countering Digital Hate, the Anti-Defamation League, and the Trust & Safety Professional Association — which functions as the credentialing body for the industry, running conferences and publishing the Journal of Online Trust & Safety. The relationship between these organizations is not conspiracy. It is networking. The same people attend the same conferences, serve on the same advisory boards, move between the same institutions, and share the same operational assumption: that the correct response to problematic speech is institutional gatekeeping at scale.

The mechanism was made concrete by the Election Integrity Partnership, which was co-led by Stanford’s Renee DiResta and established in July 2020. The EIP submitted 4,800 flagged URLs directly to social media platforms with specific action recommendations: reducing “discoverability,” “suspending accounts’ ability to tweet,” removing posts. This was not content moderation by the platforms. This was an external organization with university prestige telling platforms what to suppress — and platforms, eager to outsource the political risk of their own moderation decisions, complied.

The Twitter Files — internal documents released by journalists given access by Elon Musk in December 2022 — documented the machinery in detail. Twitter held regular meetings with the FBI, DHS, and other agencies. A formal system existed for receiving “thousands of content reports from every corner of government including HHS, Treasury, NSA, and local police, often accompanied by spreadsheets of hundreds or thousands of account names for review.” Both the Biden campaign and the Trump White House flagged tweets for removal. The system was not partisan. It was institutional.

The Virality Project — a Stanford initiative — told Twitter that “true stories that could fuel hesitancy” should be treated as “Standard Vaccine Misinformation on Your Platform.” True stories. Not false claims. True reports of blood clots from AstraZeneca vaccines, true stories of vaccine recipients who developed thrombocytopenia. The Virality Project’s final report classified it as misinformation to suggest the vaccine does not prevent transmission or that governments were planning vaccine passports — both of which turned out to be true.

Murthy v. Missouri reached the Supreme Court in June 2024. Missouri, Louisiana, and five social media users alleged that government communications urging platforms to act against COVID and election “misinformation” violated the First Amendment. The Court dismissed the case 6-3 on standing — the plaintiffs could not show government statements were the likely cause of specific platform actions against them. The Court did not rule on the merits. Justice Alito’s dissent, joined by Thomas and Gorsuch, called it “one of the most important free speech cases to reach this Court in years.” Five days later, in Moody v. NetChoice, the Court ruled unanimously that social media platforms are protected by the First Amendment when exercising editorial discretion — even through algorithms.

The infrastructure has since collapsed, or at least contracted. The Stanford Internet Observatory was reduced to three staff by June 2024 — DiResta’s contract not renewed, Stamos departed citing political pressure. The Global Engagement Center was defunded and closed on December 23, 2024. The Global Disinformation Index was defunded by both the U.S. and U.K. governments after an analysis found every one of its 10 “riskiest” outlets leaned right and all but one of its 10 “least risky” leaned left. Mark Zuckerberg announced in January 2025 that Meta would end its third-party fact-checking program, scale back automated moderation to focus on “high severity violations” only, and move its Trust & Safety teams from California to Texas.

But the people relocated. They did not retire. DiResta moved to Georgetown. Stamos remains at Stanford. Yoel Roth, Twitter’s former Head of Trust & Safety, landed at UC Berkeley and the Carnegie Endowment for International Peace. The same career path — platform, think tank, government advisory role, back to academia — persists because the institutions that sustain it persist. The conferences still run. The journal still publishes. The professional credentialing body still certifies.

The counterpoint is worth registering. The EFF — which no one accuses of being a censorship apologist — opposes both government coercion and platform over-moderation, and explicitly affirms that social media platforms have First Amendment rights to curate content. FIRE, the Foundation for Individual Rights and Expression, found that “Americans don’t trust the government to make social media content decisions” and that even when platforms’ moderation decisions are “unwise or biased,” empowering government to dictate those decisions would be “a cure far worse than the disease.” Nadine Strossen, former president of the ACLU: “No matter how great the potential harm of the speech, the potential harm of censorship is even greater.”

The infrastructure question is not whether content moderation should exist. The Something Awful model worked. Ten dollars and a ban button. Community moderator with context and judgment. The infrastructure question is what happens when the moderation function is outsourced to a revolving-door network of think tanks, government agencies, and professional organizations whose operational assumption is that they should decide what speech is acceptable at civilizational scale. The answer, as of early 2026, is that the infrastructure overreached, the political wind shifted, and the institutions are being dismantled — but the people who built them are reconstituting in new positions with the same assumptions and the same Rolodex.