Commentary accounthandle withheld
Trump announces DOJ investigation into 500,000 *illegal* mail-in ballots sent out by the blue state of Maryland. Stop the fraud!
1,944,012 interactions across three posts
Confidential · not for distribution
What people see shapes the decisions they make
The average person spends about two hours and twenty minutes a day on social media (DataReportal, 2025).
Where there's smoke… is there fire?
Accounts that look independent repeat baseless claims about elections, health, and policy until people start to doubt the facts. Fooled repeatedly, people devalue every signal.
Sources: Hoes et al., Nature Human Behaviour 2024 · Gallup, Sept 2025.
We rely on journalists to check it. They've gone dark.
What happens
events, claims, campaigns
manipulation floods in
Platforms & feeds
where people read it
Journalists & fact-checkers
the check on what's real
Communities decide
on unchecked feeds
the stopgap is gone, so manipulation reaches people directly
Accounts disappear before alerts appear
A listening tool reports the conversation after the fact. By then the accounts driving it are often suspended, renamed, or deleted.
Telling you what happened is too late. The keywords never change, so the finding arrives after the campaign has run.
Telling you as it happens enables response. Conversations rewrite the keywords, so the net widens while it is still running.
Manipulation reaches the decisions that matter
Jobs · trafficking
Fake online job ads lure workers into forced-labour scam compounds.
Health · vaccines
Anti-vaccine fear cut Samoa's vaccination rate below 30%; dozens of children died.
Elections
A coordinated TikTok operation got the election annulled; the evidence vanished.
We show journalists what's shaping your feed, before it shapes your views
What happens
events, claims, campaigns
Platforms & feeds
public posts, through platform APIs
Journalists & fact-checkers
their view, restored through Arbiter
Communities decide
informed choices
collect
hand back
Arbiter is a public-interest AI system: it maps the networks behind deception before it sways people's minds.
Search every feed, aggregate what moves together, investigate who's behind it
1 · Search
X
YouTube
Reddit
2 · Aggregate
Coordinated praise
44 posts · 3 platforms
Recycled talking points
31 posts · 2 platforms
3 · Investigate
one hub drives the rest
Arbiter is live, and already at scale
- 100+ organizations run cited investigations on one shared platform: newsrooms, fact-checkers, and civil-society groups
- Public data only, collected within platform terms of service and through platform APIs
~20B
interactions a month on the posts we collect
3.7M+
posts collected across investigations
100+
organizations, from 21 countries
318
case studies created; 31 published
Interactions = views, comments, and likes on the posts we collect. Public case studies at arbiter.simppl.org/studies.
One investigation: the fake-job scam pipeline
We asked Arbiter to map how Filipino workers get recruited abroad, and where the scams hide. Built with Migrasia, and since replicated for Indonesian workers.
1st Dynamic Personnel Resources Inc.
FacebookDEPLOYMENT UPDATE! Another milestone achieved!🥳 Congratulations to our newly deployed workers bound for Japan!🎉 …
Health Care Jobs · amplifier page
FacebookDEPLOYMENT UPDATE! Another milestone achieved!🥳 Congratulations to our newly deployed workers…
What Arbiter found
- 759 posts from 383 accounts formed 14 networks sharing phones, emails, templates
- One agency and a look-alike page co-posted 33 times, same numbers, same emails
- The displayed licence, checked against the DMW registry, had quietly expired
The Global Anti-Scam Alliance, 124 members across 90 countries, invited us to present the findings at its Bangkok summit. The report reached the Philippines Department of Migrant Workers.
Public reports: Philippines · Indonesia replication.
Monitoring tools show what trends. Arbiter shows who is behind it.
15 May
Maryland announces a printing vendor error: some voters were sent the wrong primary ballot.
22–27 May
The vendor cannot say who was affected, so replacements go to every mail-in voter, in envelopes marked “REPLACEMENT BALLOT INSIDE”. One ballot counts per voter.
Days later
Those mailings are recast online as “500,000 illegal mail-in ballots”: the same templated line from 50+ accounts across platforms.
Jared DeMarinisMaryland State Administrator of Elections
It bears repeating that no fake OR illegal mail-in ballots were distributed. Mail-in voting is not a fraud.
16,277 interactions across two posts
The claim
1,944,012 the loudest account pushing it
The correction
16,277 the state's chief election official
Sources: FactCheck.org, May 2026 · Maryland State Board of Elections. Interactions per account, 1–19 May; the study cites 469 news documents. Live study.
What the argument is for, and what Maryland did about it
The same distrust of mail voting is the argument behind the SAVE Act.
Financial news account16 May 2026 · handle withheld
TRUMP: PASS THE SAVE AMERICA ACT NOW, citing concerns over mail-in voting
111,142 interactions
2.1M
Mail-in ballot controversy
256 posts
2.4M
Legislative push for reform
141 posts
Two themes, one window. The ballot claim opened the conversation; the argument for new rules drew more than the claim itself.
The stakes · the SAVE Act
- It requires documentary proof of citizenship to register, in person
- Analysts say that would effectively end mail and online voter registration
- 21.3 million citizens lack ready access to those documents (Brennan Center)
- Passed the House twice; stalled in the Senate at the 60-vote threshold
The response
- Maryland’s election consultants found Arbiter through a partner newsroom
- In under an hour they mapped the networks pushing the SAVE Act
- The false ballot claim was the pretext those networks used
- The State Administrator of Elections signed a letter committing to county and statewide deployment
Deployment is subject to funding. Democracy International: 100+ projects worldwide, a USAID partner since 2003.
Sources: H.R. 8281 · H.R. 22 · Brennan Center · Democracy International.
An investigation that took two weeks now takes thirty minutes
Arbiter cuts the time an investigation takes, and makes it reproducible.
Before Arbiter
2 weeks
manual cross-platform triage of scam-recruitment ads, one platform and one post at a time
With Arbiter
30 min
Arbiter collects hundreds of suspect ads, maps the coordinated networks, and drafts the investigation; the analyst verifies and publishes
And it replicates. The pipeline we built for the Filipino investigation re-ran for Indonesia: new country, new language, same afternoon. Every partner we onboard makes the next one faster.
Time figures from our Migrasia engagement. Public reports: Philippines · Indonesia replication.
Six spaces we work in
Youth safety
Manosphere narratives: 3,361 posts, 127M interactions on X in two months. Child-safety early warning with Odisha state officials.
Democracy
Election monitoring with Maryland’s election consultants. We traced one false ballot claim across 50+ coordinated accounts.
Full case earlier in this deck · Democracy International
Anti-scam
Fake overseas job ads that end in trafficking. With Migrasia we mapped 14 coordination networks across 383 accounts, twice.
Climate
Evidence for advocacy groups facing organized denial. Across 645 posts, one narrative took 45% of all interactions.
DeSmog signed up on their own
Health
Antenatal and menstrual-health literacy over WhatsApp, in trials in Maharashtra and Bangladesh. We benchmark LLMs in Hindi and Marathi.
Economic mobility
We train newsroom staff with DW Akademie and JournalismAI, undergraduates through NextGenAI, and health workers. 2,200+ trained.
Across all six, we detect and combat synthetic media and AI-generated persuasive campaigns, reading cognitive and behavioral signals alongside the content itself: who the accounts are, when they post together, and how the wording repeats.
All six on one page: simppl.org/simppl-brochure.pdf
Who signed up
Rappler
dpa
Deutsche Welle
Bloomberg
Chequeado
Maldita.es
La Nación
Factchequeado
Africa Uncensored
NEST Mongolia
Jagran
DeSmog
Sakal Media
Nepal Fact Check
Migrasia
The Baltimore Banner
Ofcom
UNDP
UNHCR
Internet Archive
Columbia Journalism School
100+ organizations across 21 countries signed up on their own; we are working with a growing subset as resources allow. A global fact-checking network wants Arbiter across its whole verified membership. That network leads the field with 140+ member organizations. Client names available on request.
When partners publish, institutions act
Partners publish
19 days early
NEST caught the coordinated push behind a secret defamation bill before it surfaced publicly, reading the signal from fewer than 60 accounts.
Mongolia's only internationally accredited fact-checkers · case report
Platforms act
2 platforms
acted on evidence our partners published: Meta's Q1 2024 Bangladesh takedown, and YouTube terminating flagged channels.
Meta Adversarial Threat Report Q1 2024 · Burkina Faso case (appendix)
Governments act
5 governments
opened conversations with us about running this work in their jurisdiction: Maryland, California, New York; Odisha in India; and the Philippines' Department of Migrant Workers.
Maryland has committed to a statewide rollout; the rest are proposals and scoping conversations, July 2026
$2M raised jointly since 2021; $450K came straight to us
- Funders include Google, Mozilla, Ford, Omidyar, Wikimedia, and Columbia's Brown Institute
- The IRS took a year longer on our 501(c)(3), so partner orgs received grants for us
- Status in hand: INSEAD ($110K) and Columbia ($100K) fund us directly this year
Raised-with-collaborators total per the SimPPL Annual Report 2025; direct-receipt split confirmed by the team, July 2026.
Earned revenue reached $120K this year
- Subscriptions from larger newsrooms and civil-society clients: annual earned revenue
- Grants pay for the free access that independent journalists and fact-checkers get
- Inbound demand is real: users arriving from Dataminr and Meltwater asking to switch
Contracts this year
$178K signed across four contracts
Grants, separate
INSEAD $110K
Columbia · Brown Institute $100K
$120K counts earned revenue this year; multi-year contract values run higher. INSEAD and Columbia are grants, listed separately.
Five-year plan: deepen journalism, scale civil society, sign agencies
2026
2027
2028
2029
2030
Journalism
DW Akademie + Indicator readers
Global fact-checking network
60 flagship newsroom institutions
Civil society
NEST, Migrasia, UNDP cohorts
Odisha child-safety early warning
State election + AG programs
Measurement of Effects for Interventions
Commercial · agencies
Global marketing agency pilot
3-4 enterprise deals, $300K earned
Agency + brand accounts scale
live today (July 2026)
proposal or scoping, dated 2026
modeled target
Grants fund journalist access; contracts pay for programs and measurement.
Two customers: one does the work, one pays for it
Impact
Fact-checkers and newsrooms
Free seats. They run the investigations and publish what we find, in their own language and their own byline.
pays for
those seats
those seats
Paying
Agencies, civil society, government
About $2,500 to $15,000 a month, signed by a director. We are in conversation with a global marketing agency, a state attorney general’s office, a state elections agency, and civil-society research organizations.
The commercial clients listed are prospects with ongoing conversations. Client names available on request.
Each customer opens the next one up
Public interest
Journalists, free seats
Networks distribute us: a global fact-checking network’s members, and Indicator.media’s 14,000 readers.
Media that pays
We measure a newsroom’s own mentions, so audience teams renew annually.
Civil society
Extremism researchers already license data. We sell the analytics layer on top.
Government
The setup we build for one agency becomes the next agency’s starting point.
Commercial, which funds the free seats
Brand marketing client
A consumer brand, and where we learned audience analytics.
Financial markets transparency project · $60K signed
A prediction market. It prices public response into its odds.
Global marketing agency
A global agency, and a channel into every brand it serves.
The two sides run on one platform. The public-interest side finds the problems and wins the grants; the commercial side pays for the free seats underneath it.
The global fact-checking network is a leading network of 140+ member organizations. Commercial clients listed are prospects with ongoing conversations; the transparency project is signed. Client names available on request.
Our first five years
2021
Founded out of NYC Media Lab; bootstrapped nonprofit
2023
First in India to win responsible AI education awards from both Google and Mozilla, then Wikimedia; 5 peer-reviewed papers published
2024
CrowdTangle shuts down; members of its own team now help us, an early vote of confidence. Meta's Q1 Bangladesh takedown follows our partners' reporting; we join the MIT venture accelerator
2025
Arbiter ships: UNDP AI cohort; first recurring revenue, $15K a quarter. Nobel Laureate Maria Ressa's team onboards
2026
Scale: Fast Forward accelerator; 100+ orgs, 21 countries, 318 case studies created; earned revenue reaches $120K. Stanford picks Arbiter for its platform workshop, a slot Meta, TikTok, and YouTube held in past years. 4 state governments reach out to partner
Funded by Google, Mozilla, Ford, Omidyar, Wikimedia, and the Brown Institute at Columbia.
The team, and the board behind it
Swapneel Mehta · Founder & Executive Director
Former postdoc at MIT and BU in AI safety. Integrity Institute Board Member, advising EU, UK, and LatAm policymakers on online safety.
Dhara Mungra · Co-Founder, leads engineering
NYU MS; former Data Scientist III at Bombora; CAIDP & U. Mannheim fellow; black belt and marathoner
Six full-time engineers build and run the platform: retrieval, agents, evaluation, infrastructure
Board of directors
Mukund Sudarshan · Anthropic
Christian Schroeder de Witt · University of Oxford
Chirag Raman · TU Delft
Sandra Khalil · All Tech Is Human
Rebecca Harvey · UK FCDO
Karan Dhabalia · entrepreneur (exited to Stripe)
Mansi Panchamia · World Bank
Stacey Peters · ex-VTDigger
Full board and team: simppl.org/team
The product is complemented by training programs we have run for years
For four to eight years we have trained journalists, students, and researchers to use these methods and tools. The programs build the audience and the credibility the product stands on.
Students · fellowship
NextGenAI. A year-long capstone fellowship where undergraduate teams build and ship AI responsibly. Sakhi, our health assistant, spun out of it.
Journalists · newsrooms
AI for Journalism. Capacity-building for newsroom leaders and reporters, run with DW Akademie and JournalismAI.
Researchers · health workers
AI for Healthcare. Training frontline health workers to run Sakhi, our WhatsApp health assistant, in local languages.
2,200+ trained across these programs. Alumni have gone on to Ivy League PhD programs, the Microsoft Research Fellowship, FAANG companies, and prominent startups.
Awards from Google and Mozilla. Programs backed by Genentech, DeepMind (via the NYU AI School), the University of Mannheim, DW Akademie, JournalismAI, and MIT. simppl.org/programs · simppl.org/testimonials
Areas of funding
Annual support, at whichever level fits your program.
$250K a year
One issue area
- Our work in any one of the six spaces
$500K a year
A global fact-checking network equipped
- Access for its 140+ verified member organizations, with training in partner countries
$1M a year
All six areas
- General operating support across all six spaces for a year
Pilot grants from $25K to $250K fund one specific program. Tell us which space matters most to you and we will align the scope with your priorities.
Budgets and milestones are ready for whichever scope you want to discuss. team@simppl.org
Three things to take with you
1
The journalists and fact-checkers communities trust lost their view of the feeds; Arbiter restores it, with every claim cited.
2
It already works at scale: 100+ organizations in 21 countries, and platforms and officials act on what they publish.
3
Support runs from a $25K pilot to $1M a year of general operating support across all six areas.
team@simppl.org simppl.org · arbiter.simppl.org · a U.S. 501(c)(3) nonprofit
Take these with you
3-minute pitch video
Fast Forward demo day: youtube.com/watch?v=Q26vMDKl5FE
One-page brochure
simppl.org/simppl-brochure.pdf: the case studies and stats on one page
Public case-study library
31 investigations, each traceable to source: arbiter.simppl.org/studies
Technical deck
The full technical walkthrough, with the climate-skepticism example: mehtaver.se/slides/arbiter.html
Methods & AI footprint
Token-efficient models that cut the climate cost of large-scale analysis: mehtaver.se/slides/css-methods
Proposal & business plan
On request: team@simppl.org