Information insecurity is increasingly treated as a security problem, not a media problem. Mirror Protocol applies creation-time provenance, portable attribution, and exportable audit trails to the newsroom — the layer where a claim is either traceable to its origin or it is not.
It does not certify that a story is true. It makes the process behind the story verifiable, so the burden of explanation shifts to whoever disputes it.
The strongest published result for after-the-fact correction is modest: a randomised trial of AFP's fact-checking reduced circulation of the checked disinformation on the largest social platform by roughly eight per cent. That is a real effect, and worth funding. It is also an eight per cent effect against a problem estimated to cost societies somewhere between US$355 and $516 billion a year.
Moderation is a downstream response. By the time a falsehood is checked, it has already been read, shared, and in many cases believed.
Governance at the creation layer rather than the moderation layer — the same argument Mirror Protocol makes for policy, for creators, and now for the press.
Sources for every figure on this page →These steps are a sequence because custody is a sequence. Each entry is written as the work is done — not reconstructed afterward, when reconstruction is exactly what is in doubt.
The brief, the question being asked, the reporter of record, the timestamp. In a dispute, the earliest defensible fact a journalist can produce is what they set out to do and when. Under the protocol that is a ledger line, not a memory or a lost message thread.
Transcription of an hour of tape. Translation of a source's testimony. Background retrieval. A model that helps restructure a lede. Normally all of that dissolves by publication and no one can say which sentence came from the recorder and which came from the reporter's judgment. The attribution ledger holds each contribution with a timestamp, a role, and a system identifier.
The council model binds each contributor to a defined role — research retrieves, synthesis drafts, verification checks. It maps onto the wall a newsroom already keeps between reporting and editing, and it means a hallucinated citation can be traced to the step that produced it instead of contaminating the whole file with doubt.
Cross-verification flags provenance gaps before release: a quote with no source log, a statistic that entered the draft without a retrieval record, a claim that changed between versions without an explanation. This is not a truth machine. It is a gate against the failure mode that actually ends careers.
A specific accounting of which tools were used and where — the difference between a disclosure that invites suspicion and one that closes it.
When a government demands a correction, when a plaintiff files, when a platform deranks the piece as synthetic, when an altered version circulates under the reporter's byline — the newsroom holds a timestamped, machine-readable record of the original and every edit that produced it.
Misinformation is rarely invented from nothing. The more effective attacks take real material — a real byline, a real quote, a real document, a real outlet's typeface — and move one element out of place. The story survives. The trust does not.
The forged byline. A fabricated article circulates under a working reporter's name, styled like their outlet, quoting sources they never spoke to.
No ledger exists for it. Survey evidence shows that when people are unsure whether an online claim is true, the thing they check against is a news outlet they trust — ranked above official government sources. Forging a byline is therefore not identity theft against one reporter; it is an attack on the public's last working verification instrument.
The doctored quote. A published piece is copied, three paragraphs are altered, and the altered version spreads faster than the correction.
Immutability under edit. The audit package holds the released version and every prior revision. Any change made outside the chain is visible as a change, and the newsroom can show precisely which words were moved.
The liar's dividend. As synthetic media becomes ordinary, a genuine recording or document can simply be dismissed as AI-generated. The accusation costs nothing to make.
Provenance presence. Authentic material arrives with capture metadata, custody, and verification checkpoints attached. The denial now has to explain a record, which is a far more expensive thing to do than to shout.
The poisoned source. Fabricated studies, invented statistics, and synthetic documents are seeded upstream, waiting to be retrieved by a rushed reporter or a research model.
Every figure carries a retrieval record naming where it came from and which system fetched it. Claims without a traceable origin are flagged at the verification gate rather than discovered by a reader.
The context strip. Real footage, real audio, real numbers — presented with the date, the place, or the surrounding question removed.
The directional vector. The ledger keeps not only what the material is, but who moved it and toward what. A stripped clip has no origin entry; the original does, with its context intact.
The volume attack. Thousands of near-identical synthetic articles bury a piece of reporting until no reader can find the original.
Machine-readable provenance gives platforms, aggregators, and archives a basis for ranking verified origin over synthetic volume — a filter that does not depend on judging content, only custody.
Every figure below is drawn from The Value of Journalism, a 2026 evidence brief by Professor Mel Bunce and Dr Beth Pearson of City St George's, University of London, commissioned by DW Akademie, IFPIM and UNESCO. Full attribution and the authors' own caveats are on the evidence page.
Estimated annual global cost of disinformation across financial, social and political markets.
Bunce & Pearson 2026, citing Sopra Steria 2026
Rank of mis- and disinformation among global risks in the World Economic Forum's 2026 Global Risks Perception Survey.
Bunce & Pearson 2026, citing WEF 2026
Reduction in circulation of checked disinformation achieved by AFP fact-checking on the largest social platform — the ceiling on correction as a strategy.
Bunce & Pearson 2026, citing Cagé et al 2025
Share of global GDP — roughly fifteen days of world military spending — that experts estimate could fund healthy public media and secure information environments worldwide.
Bunce & Pearson 2026, citing Forum for Information and Democracy 2021
Per-capita increase in average municipal deficit associated with newspaper closure in the United States. Information infrastructure is economic infrastructure.
Bunce & Pearson 2026, citing Gao et al 2020
Associated reduction in real GDP growth when a country declines in press freedom — slow to recover.
Bunce & Pearson 2026, citing Nguyen et al 2021
This has to be said before anyone else says it. Source protection is the first duty of the craft, and a naive provenance system is a subpoena waiting to be served. The newsroom implementation therefore separates two things that a general-purpose ledger keeps together.
A source entry can be recorded as a verified but sealed contribution: the chain proves that a human source of a stated type was present, contributed at a stated time, and was verified by a named editor — without carrying the identity in the exportable record at all.
The evidence bundle belongs to the outlet. A protocol that parks a reporter's custody log on a third party's servers has not protected the reporter; it has relocated the risk.
When countries experience democratic backsliding, media freedom is frequently the first element of democracy to come under attack. Any system that decides which work carries an authentic mark is therefore a target — and a provenance registry controlled by a state does not merely fail, it launders authority.
Verification must be reproducible by any party holding the record, not granted by one body that can be pressured, purchased, or replaced.
The same record must yield the same result regardless of who checks it. A test that returns different answers for different examiners is a permission system wearing a verification costume.
Work without a chain must remain publishable and readable. The moment missing provenance becomes grounds for suppression, the protocol has been turned into a censorship instrument.
The literature on journalism's impact is heavily weighted toward the United States and Europe, and researchers working on disinformation warn that interventions ignoring local dynamics are frequently ineffective and can cause harm. One study of conflict-affected communities in the Central African Republic found that distrust can play a protective role — where institutions have been predatory, refusing to believe them is a survival skill, not a deficit.
A provenance system says, in effect: trust what carries this mark. Mirror Protocol's answer is not to argue the point but to design for it — verification any party can perform independently, custody held by the newsroom rather than a central authority, and pilots built with local newsrooms rather than delivered to them. The architecture is modelled on council governance drawn from Nupe and Ibibio oral tradition — documented consensus among accountable roles — rather than on a single adjudicating authority.
Attribution travels with the work across platforms and vendors. A freelancer who filed to an outlet that later folded holds portable proof of contribution. The mechanism that protects a session musician's split protects a stringer's byline.
The contributors most often absorbed into someone else's credit are logged by role at the moment they contribute — a labour protection as much as an editorial one.
Small newsrooms carry the same legal exposure as large ones with none of the counsel. The news-desert research shows what their absence costs a community in deficits, corporate misconduct and pollution.
Honesty about the ceiling is the reason to trust the floor.
A twenty-minute briefing covers the three artifacts a newsroom would receive — disclosure pack, attribution ledger, audit export — and a redacted sample of each against a real story workflow.