ELECTION INTEGRITY PARTNERSHIP
- Status
- WOUND DOWN — 2020 election-rumor flagging consortium (Stanford Internet Observatory, UW Center for an Informed Public, Graphika, Atlantic Council DFRLab); mechanism persists via the Virality Project and DSA trusted-flagger architecture
- Hazard — Reach
- 78
- RCH / FND / ENT
- 7 / 6 / 7
- Conduct
- PROXY-ROUTING — THE DENIABLE FLAG PIPELINE
Institutional Archetype
THE TICKET QUEUE — The Election Integrity Partnership was a temporary swarm: four institutions that converged for a single election season, built one shared piece of software, and dispersed. The software is the archetype. The EIP ran a Jira ticketing system — the same category of project-management tool used to track bug reports — repurposed to track claims about an election. “External stakeholders,” federal agencies among them, submitted reports; analysts searched the platforms for matching content; the significant tickets went to Facebook, Instagram, Twitter, TikTok, and YouTube “with specific recommendations.” No agency issued an order and no platform received a command. A ticket simply appeared in a queue, and a queue is not a censor — it sits, patiently, between the people who want speech gone and the people with the button. This profile scores that mechanism — the deniable seam between government-adjacent flaggers and platform enforcement — not the researchers who staffed it, who were themselves harassment and subpoena targets.
Mandate & Origin
The EIP was stood up in summer 2020, “100 days before Election Day,” by four organizations: the Stanford Internet Observatory (lead), the University of Washington Center for an Informed Public, Graphika, and the Atlantic Council’s Digital Forensic Research Lab (DFRLab). Roughly 120 people worked across the four. Its stated mission was to “identify mis- and disinformation before it went viral and during viral outbreaks,” share “counter messaging,” and increase “understanding of the dynamics shaping the information space.” It was a consortium, not an institution — a load-bearing distinction: the charges here are the swarm’s conduct, not a fresh charge sheet against any member.
Funding & Backers
The EIP had no separate budget diagram of its own to publish; it was staffed and resourced by the four member institutions and their existing philanthropic and university funding, which is why the consortium-versus-member distinction matters. The relevant “backers,” in the mechanism sense, are the intake side of the queue: the “external stakeholders” permitted to feed tickets in — including federal agencies and government-funded organizations. The House Judiciary Committee characterizes the arrangement as one created “at the request” of the Cybersecurity and Infrastructure Security Agency, working “directly with the Department of Homeland Security and the Global Engagement Center,” and providing “a way for the federal government to launder its censorship activities.” That is the committee’s characterization, footnoted to an internal Atlantic Council email; the EIP disputes it directly (see below). The funding note that survives without dispute is structural: a pipeline that government could feed but did not have to be seen operating.
Actions & Leadership Choices
The mechanism, judged on the record. By the EIP’s own accounting in its 290-page final report, The Long Fuse, it processed 639 in-scope tickets, and of the links it shared with platforms, 35% were “labeled, removed, or soft-blocked.” An outcome rate high enough that the pipeline plainly moved the needle. In one internal note quoted by the committee, the EIP reported that after it flagged content, “Twitter took action almost immediately.” The House committee later obtained data for “almost 400 EIP Jira tickets” — and states it did so only after Stanford produced the records “under the threat of contempt of Congress.” The compulsion cuts two ways: to the committee, resistance to disclosure reads as consciousness of guilt; to the researchers, a contempt-backed subpoena is what it looks like when the state turns on people who studied speech. The record notes only that the data came out the hard way.
The defense, at full strength. The EIP’s position, in its own words: it had no power to remove anything — “social-media platforms, not EIP, decided which action to take” — and it denies acting as a government cutout. Both national party committees were invited to submit tickets; the DNC submitted four and the RNC submitted none. A Poynter review found left-leaning accounts (Occupy Democrats among them) were also flagged, and the EIP’s partnership with CISA “began under the Trump administration” and was “reviewed and approved by Trump Administration attorneys.” No court has adjudicated the EIP’s conduct unlawful. The one case to reach the Supreme Court, Murthy v. Missouri, was decided 6–3 on standing — the merits of the government-pressure question were never reached.
The mechanism outlived the swarm. The EIP did not reconvene at 2020 scale, but the design propagated. Its COVID-era successor, the Virality Project, ran a Stanford-led consortium coordinating across YouTube, Twitter, Facebook, Instagram, Medium, TikTok, and Pinterest; per the Twitter Files (Part 19) and reporting in Reason, it recommended platforms flag material including true reports of vaccine side effects — a shift the reporting characterized as moving from judging truth to policing narrative. Abroad, the same trusted-flagger architecture is now written into the EU’s Digital Services Act. The related dashboard-and-attribution model appears again in Hamilton 68 and in the flagging work of Graphika and the Stanford Internet Observatory. The specimen dies; the burrow gets reused.
CONDUCT verdict: PROXY-ROUTING — THE DENIABLE FLAG PIPELINE. The documented mechanism — a queue that government-adjacent stakeholders could feed, routing flags with recommended enforcement to five platforms at a self-reported 35% action rate — is settled, largely by the coalition’s own records. The characterization that this was “censorship by proxy, ordered by the state” is the House committee’s, disputed on specifics by the EIP, and never adjudicated. The hazard is the pipeline, not the researchers: a reproducible, deniable seam that no single actor need be seen operating, prototyped by academics and since propagated into a COVID successor and into law on two continents.
Sources: House Judiciary Committee — EIP Jira-ticket staff report (Nov. 6, 2023); House Judiciary Committee — “The Weaponization of CISA” staff report (June 26, 2023); Murthy v. Missouri, No. 23-411 (U.S. 2024) — opinion (decided on standing); The Election Integrity Partnership — The Long Fuse: Misinformation and the 2020 Election (final report); The Election Integrity Partnership — “A Statement from the Election Integrity Partnership”; UW Center for an Informed Public — final-report announcement (Mar. 2, 2021); Atlantic Council — “The Long Fuse” report page; Poynter — “Election Integrity Partnership targeted election misinformation, not conservatives” (2022); The Seattle Times — “Attacked by conservatives, UW misinformation researcher gears up for 2024”; Reason — “Stanford Researchers Lobbied Twitter to Treat COVID-19 Facts as Misinformation” (Mar. 17, 2023); Twitter Files, Part 19 — the Virality Project (Substack).
Get updates on the Evil Robots series
Newsletter essays on AI escape, deception, and the humans who built them.