FACT
Directly sourced
A claim stated by a cited source, filing, register, decision, court record, or official page.
Methodology
A reader's guide to the dossier's evidence rules: what is directly sourced, what is analytical inference, what is only structural proximity, and which tempting claims have been rejected.
The dossier separates what a source directly says from what the dossier concludes from multiple records. A sourced record is marked FACT. A connection drawn by this investigation is marked INFERENCE.
The strongest thesis row is commercial capture: Thorn sells Safer, reports Safer/program-service revenue, and lobbied the EU scanning file. The page does not convert that into proof of a covert US-government command chain.
"No evidence found" is not treated as "false." It means the public record checked did not support the claim. When a claim is false, stale, overstated, or needs precision, it belongs in the ledger below.
FACT
A claim stated by a cited source, filing, register, decision, court record, or official page.
INFERENCE
A conclusion the dossier draws from the record. It may be strong, but it is still ours.
Thesis claims use verdict chips; ledger claims use rejection chips.
PROVEN
Direct records support the claim as stated.
STRUCTURAL
The record shows proximity, incentives, or lineage, not a directing hand.
DOCUMENTED
Dependency or architecture is documented; intent may remain inference.
SPLIT
One part is proven while another part remains unproven.
SPECULATIVE
Retained only as a weak hypothesis, not as evidence.
Loaded from the Wave 1 data contract with verdict distinctions intact.
Sell surveillance software
PROVEN
Thorn sells Safer, Safer/program-service revenue is 49% of FY2024 revenue, and Thorn lobbied for EU scanning mandates.
Commercial capture is proven; covert US-government direction is not.
The masking mechanism
PROVEN
Child-protection framing lowered scrutiny through emotional framing, survivor-pairing strategy, and illegal DG HOME microtargeting.
Keep the sourced scrutiny effects separate from unsourced claims about secret direction.
US-intelligence ties
STRUCTURAL
Digital Reasoning, Palantir, Lonsdale, and related interlocks document proximity and lineage.
No documented US-government cash into Thorn and no proven covert command.
Exploitable security holes
SPLIT
Mandated backdoors and client-side scanning risks are documented technically.
Intent by US actors to exploit those holes remains unproven.
Undermine EU sovereignty
DOCUMENTED
Dependency on US-seeded tooling, hash corpora, definitions, and enforcement architecture is documented.
Deliberate sovereignty-undermining remains inference.
Industrial/economic espionage
SPECULATIVE
No evidence Thorn or Chat Control is a vehicle for US industrial espionage.
Retain only as a weak hypothesis and do not promote Tier 3 comparisons into proof.
Corrections and challenges
Rejected claims are visible, including claims that would have helped the thesis.
FALSE / UNPROVEN
Corrected wording
No primary source documents a direct transfer; use structural proximity language.
Why this is bounded
This would strongly support the thesis, but the checked record supports technology lineage, coalition overlap, and revolving-door proximity rather than a direct covert funding chain.
Evidence status
FACT Funding / thesis-favorable rejected claim
NEEDS-PRECISION / STALE
Corrected wording
Lonsdale sat on Thorn's board from 2018 through February 2022; do not state a current seat.
Why this is bounded
The board link is real and important for the drafting run-up, but current-seat wording is stale and would overstate the present tie.
Evidence status
FACT People board / thesis-favorable rejected claim
NEEDS-PRECISION / STALE
Corrected wording
Ruiz Perez sat on Thorn's board until he stepped down in January 2024; use past-tense wording.
Why this is bounded
The Europol-to-Thorn board move is documented, but present-tense phrasing is wrong after the January 2024 departure.
Evidence status
FACT People board / thesis-favorable rejected claim
FALSE / UNRESOLVED
Corrected wording
Do not equate Fede/Federico with Fernando Ruiz Perez; the open-source referent is unresolved.
Why this is bounded
The nickname match failed and no distinct Federico was found in the checked Thorn, WeProtect, Europol, or Ombudsman records.
Evidence status
FACT People board
OVERSTATED
Corrected wording
John F. Bennett is a retired FBI Assistant Director, not the head of the FBI; Ernie Allen is NCMEC/WeProtect, not FBI.
Why this is bounded
The law-enforcement board tie is real, but the title inflation would make a documented proximity record look like something stronger.
Evidence status
FACT People board / thesis-favorable rejected claim
OVERSTATED
Corrected wording
Spotlight was co-built with Digital Reasoning, a documented In-Q-Tel portfolio company; that is a historical product tie, not proof the CIA built it.
Why this is bounded
The correct chain is Spotlight to Digital Reasoning to In-Q-Tel. Compressing that into "the CIA built it" silently upgrades proximity into direction.
Evidence status
FACT Technology intelligence / thesis-favorable rejected claim
NEEDS-PRECISION / STALE
Corrected wording
Digital Reasoning's In-Q-Tel/Spotlight tie is historical, running 2011-2020, and Thorn transitioned Spotlight out of Thorn in May 2024.
Why this is bounded
The product tie is among the strongest intelligence-adjacent records, but present-tense wording is no longer supported.
Evidence status
FACT Technology intelligence / thesis-favorable rejected claim
NEEDS-PRECISION / STALE
Corrected wording
Palantir is current in WeProtect; Thorn Task Force membership is historical and not verifiable as current.
Why this is bounded
The WeProtect membership is current and citable; current Thorn Task Force membership cannot be confirmed because Thorn no longer publishes that member list.
Evidence status
FACT Technology intelligence / thesis-favorable rejected claim
FALSE / UNPROVEN
Corrected wording
No credible source supports the Thorn-to-Hacking-Team or malware claim; do not use it.
Why this is bounded
This is a thesis-friendly claim that failed the credibility threshold and should not be re-imported.
Evidence status
FACT Technology intelligence / thesis-favorable rejected claim
FALSE
Corrected wording
The claim is not in Google's grantee lists; do not cite it as a Thorn funding source.
Why this is bounded
Thorn has other documented Big Tech ties, so the unsupported Google AI Impact Challenge claim is unnecessary and misleading.
Evidence status
FACT Funding / thesis-favorable rejected claim
FALSE
Corrected wording
Childlight is funded by the Human Dignity Foundation, not CIFF.
Why this is bounded
The funding map must not invent funder links just because CIFF appears elsewhere in the child-safety ecosystem.
Evidence status
FACT Funding
NEEDS-PRECISION
Corrected wording
ECLAG has no legal entity; Oak channelled $24M+ since 2019 to member NGOs building the campaign, not to ECLAG as a recipient.
Why this is bounded
The money pattern is documented, but the recipient has to be described precisely so the public funding surface remains reusable.
Evidence status
FACT Funding / thesis-favorable rejected claim
NEEDS-PRECISION
Corrected wording
The documented vendor revolving door is exactly two named Europol-to-Thorn moves, plus retired FBI Assistant Director John Bennett on the board.
Why this is bounded
The sweep did not widen the pattern through Heat Initiative or WeProtect; adjacent moves should stay separate from the vendor-capture claim.
Evidence status
FACT People board / thesis-favorable rejected claim
FALSE
Corrected wording
Use 311 for rejection, 228 against, 92 abstentions for the March 2026 plenary rejection.
Why this is bounded
The larger false margin would overstate parliamentary opposition; the correct tally is still materially important.
Evidence status
FACT Vote counts / thesis-favorable rejected claim
FALSE / INVERTED
Corrected wording
The checked record says 322 for and 255 against; it failed because it missed the 361 absolute-majority threshold.
Why this is bounded
The correction strengthens the critique: a majority backed narrowing the regime but could not clear the procedural threshold.
Evidence status
FACT Vote counts / thesis-favorable rejected claim
NEEDS-PRECISION
Corrected wording
The adopted 26 Nov 2025 detection-free general approach is Council doc 15318/25; national positions still rest on leaked WK-series cables.
Why this is bounded
This keeps public Council texts and leaked national-position evidence in their proper lanes.
Evidence status
FACT Vote counts
NEEDS-PRECISION
Corrected wording
For the 9th-term file, the ECR shadow was Vincenzo Sofo and the ID shadow was Annalisa Tardino; the 51-2-1 tally is confirmed.
Why this is bounded
Shadow names and dissenters matter for later political maps, so this correction is part of the baseline credibility apparatus.
Evidence status
FACT Vote counts
NEEDS-PRECISION
Corrected wording
The coalition agreement does not contain the word Chatkontrolle; the warrantless-monitoring taboo line is a ministerial statement, not coalition text.
Why this is bounded
Germany's opposition record remains important, but the source attribution must be exact.
Evidence status
FACT Vote counts / thesis-favorable rejected claim
NEEDS-PRECISION
Corrected wording
Use seven named states; the European Service Network agency attribution is unconfirmed in the opened primary record.
Why this is bounded
The illegal microtargeting finding stands, but state count and agency attribution must not be padded.
Evidence status
FACT Campaign mechanics / thesis-favorable rejected claim
NEEDS-PRECISION / OVERSTATED
Corrected wording
As of the checked record, it is a Commission proposal; kill switch is commentator gloss, not statutory language.
Why this is bounded
The sovereignty discussion can cite the proposal's assurance-level and operational-discontinuity language without turning it into enacted law.
Evidence status
FACT Sovereignty / thesis-favorable rejected claim
OVERSTATED
Corrected wording
Ackerman is the lone square holding; Miller declined to decide and Wilson did not reach the issue. The real split concerns private-search/hash-match doctrine.
Why this is bounded
This keeps the NCMEC chokepoint claim legally precise and prevents a stronger-but-wrong circuit-split framing.
Evidence status
FACT Sovereignty / thesis-favorable rejected claim
NEEDS-PRECISION
Corrected wording
Say 99,375 reports, or 48.3%, were deemed criminally irrelevant; a report is not the same as a person.
Why this is bounded
The error-rate point remains useful for technical risk, but overstating the unit undermines credibility.
Evidence status
FACT Technology / thesis-favorable rejected claim
NEEDS-PRECISION
Corrected wording
The phrase applies to new material and grooming; known-material scanning gets narrower proportionality concerns, not the same fail verdict.
Why this is bounded
This preserves the technical/legal critique while keeping the EPRS quote within its actual scope.
Evidence status
FACT Technology / thesis-favorable rejected claim
NEEDS-PRECISION / STALE
Corrected wording
Describe the roadmap as not yet published or overdue as of the checked date, not as a still-pending future target.
Why this is bounded
The Q2 2026 target window closed on 30 June 2026, and the checked Commission pages still used future-tense roadmap language.
Evidence status
FACT Going dark / thesis-favorable rejected claim
UNRESOLVED / NEEDS-PRECISION
Corrected wording
Keep the descriptor former FBI official, but do not attach a name.
Why this is bounded
The open record checked carries the quote and institutional background, not the person's identity.
Evidence status
FACT Going dark
FALSE
Corrected wording
Use Europe's techno-borders for the Statewatch and EuroMed Rights report.
Why this is bounded
The cited border-tech figures trace to the 2023 Statewatch report; no report with the circulated title was found in the checked record.
Evidence status
FACT Industry
FALSE
Corrected wording
Do not list Atos as an EOS member; Idemia is the relevant confirmed French member.
Why this is bounded
The checked EOS member page does not include Atos, so the industry map must not inflate membership links.
Evidence status
FACT Industry / thesis-favorable rejected claim
NEEDS-PRECISION
Corrected wording
Treat about EUR300m as a eu-LISA facial-recognition programme budget line, not as the disclosed sBMS contract value.
Why this is bounded
The award notice names the vendors and duration but does not disclose that euro figure for the contract.
Evidence status
FACT Industry / thesis-favorable rejected claim
NEEDS-PRECISION / UNRESOLVED
Corrected wording
Do not assert a Thales-eu-LISA contract value unless a specific public award is pinned.
Why this is bounded
Thales has documented EU border-security roles, but the checked central biometric contract was Idemia and Sopra Steria, not Thales.
Evidence status
FACT Industry / thesis-favorable rejected claim
NEEDS-PRECISION
Corrected wording
Use EUR461m over 2019-2027 for the five interoperability components only.
Why this is bounded
The pinned figure is useful, but it must not be blended with separately budgeted border-system builds.
Evidence status
FACT Industry
NEEDS-PRECISION
Corrected wording
Use May 2024 for the 42 recommendations and November 2024 for the concluding report.
Why this is bounded
The date fix prevents later timeline work from shifting the lawful-access chain by a year.
Evidence status
FACT Going dark
NEEDS-PRECISION
Corrected wording
Use approximately GBP425,000 or over GBP400,000, depending on the cited source.
Why this is bounded
The figure is comparative Tier 3 context and must not be made sharper than the checked record supports.
Evidence status
FACT Foreign influence
NEEDS-PRECISION
Corrected wording
Attribute the Swedish file ownership to Justice Minister Gunnar Strommer.
Why this is bounded
The Swedish Council-position surface needs the responsible office correct before later timeline or people work builds on it.
Evidence status
FACT Sweden
NEEDS-PRECISION
Corrected wording
Describe Sweden as drifting, including neutrality during its own 2023 Council presidency.
Why this is bounded
A flat support label erases the timeline variation that the Swedish record actually shows.
Evidence status
FACT Sweden / thesis-favorable rejected claim
NEEDS-PRECISION
Corrected wording
Treat the bloc as varying by vote and include MP for the September 2024 vote.
Why this is bounded
This keeps future Swedish vote cards from flattening party positions into a single stable bloc.
Evidence status
FACT Sweden
NEEDS-PRECISION / STALE
Corrected wording
Use 3 November 2025 for the Dagens ETC publication date.
Why this is bounded
The date is now pinned and should not remain hedged in timeline work.
Evidence status
FACT Sweden
NEEDS-PRECISION
Corrected wording
Use Palantir EVP Noam Perski; Dagens ETC names Anders Friden as Sweden chief.
Why this is bounded
The people surface must keep title and meeting-participant details exact.
Evidence status
FACT Sweden
NEEDS-PRECISION
Corrected wording
Use Dagens ETC's exact print form, Demokratiska pa alla satt, when quoting the headline.
Why this is bounded
This is a quote-form correction, not a change to the substantive Palantir/Acus finding.
Evidence status
FACT Sweden
NEEDS-PRECISION
Corrected wording
Record the two-person debate, the minister's procurement-law answer, and that it closed with no vote or committee follow-on.
Why this is bounded
The outcome is now logged, so later Swedish accountability work should not treat it as unresolved.
Evidence status
FACT Sweden
UNPROVEN
Corrected wording
Keep ECHELON interception and diplomatic intervention as historical context; do not assert a direct intelligence-to-company transfer.
Why this is bounded
This is the clearest thesis-friendly industrial-espionage claim the publication plan says to reject. The EP record found no substantiated direct hand-off.
Evidence status
FACT Industrial espionage / thesis-favorable rejected claim
REFUTED
Corrected wording
Use Cambridge Analytica only for documented offers and methods, not proven electoral impact.
Why this is bounded
The claim is comparative context, not core Thorn evidence, and the checked process did not substantiate vote-changing impact.
Evidence status
FACT Comparative
UNVERIFIED
Corrected wording
Drop the Iran 80-90% line and keep the documented Crypto AG figures in their proper scope.
Why this is bounded
The exact statistic was not verified and should not be imported as state-through-company proof.
Evidence status
FACT Comparative / thesis-favorable rejected claim
FALSE
Corrected wording
The 2023:560 FDI regime does cover indirect investments; only the no-retroactivity point stands.
Why this is bounded
This keeps comparative foreign-investment material from overstating a Swedish legal gap.
Evidence status
FACT Comparative
STALE
Corrected wording
Drop Meng after the deferred-prosecution deal and December 2022 dismissal; the corporate case continues.
Why this is bounded
This prevents stale person-level claims from contaminating comparative state-company examples.
Evidence status
FACT Comparative
OVERSTATED
Corrected wording
Treat the tie as illustrative, not as a proven causal chain.
Why this is bounded
The example may illuminate data-breach risk, but the checked record does not prove causation.
Evidence status
FACT Comparative
UNSOURCED
Corrected wording
Drop the 469 figure unless it is pinned to a source; keep separately confirmed VPN-removal records.
Why this is bounded
The VPN comparison is Tier 3 context and should not carry unsourced numeric precision.
Evidence status
FACT Vpn
CORRECTED
Corrected wording
Wisconsin's VPN provision was removed; Michigan HB 4938 is the VPN bill to cite.
Why this is bounded
This keeps US-state VPN references current and avoids overstating the Wisconsin bill.
Evidence status
FACT Vpn
NEEDS-PRECISION
Corrected wording
Use about 0.8 percentage points higher growth per 10 percentage points of trust, or about 0.56 standard deviations of growth per one standard deviation of trust.
Why this is bounded
The original wording garbled a standardized coefficient into percentage points.
Evidence status
FACT Economic framing
RE-SOURCE FLAG WITHDRAWN
Corrected wording
The 45% figure is genuinely reported in the 2022 paper, scoped to Swiss internet users and opinion sharing.
Why this is bounded
The correction narrows the figure and removes the earlier reattribution warning.
Evidence status
FACT Economic framing
NEEDS-PRECISION
Corrected wording
Cite the paper for reduced mayoral electoral competition, not corruption; the second author is Jay T. Jennings.
Why this is bounded
The paper supports an accountability argument, but not the corruption claim as stated.
Evidence status
FACT Economic framing
NEEDS-PRECISION
Corrected wording
State that least-corrupt governments collect about 4% of GDP more in tax than most-corrupt peers at similar income levels.
Why this is bounded
The IMF figure is a comparative revenue gap, not a direct absolute loss estimate.
Evidence status
FACT Economic framing
CORRECTED
Corrected wording
Use 2% in 2010 for population reporting a bribe, and 1% for road police in the Life in Transition Survey.
Why this is bounded
The anti-corruption example is confirmed, but the circulated year and percentage are not supported.
Evidence status
FACT Economic framing
CORRECTED / GARBLED
Corrected wording
Use HRW's account that Rwanda removed some 500 judges and appointed 223, not all 503.
Why this is bounded
The reform example can be cited only with the corrected denominator and wording.
Evidence status
FACT Economic framing
CORRECTED
Corrected wording
Use the DCEC creation as confirmed, but do not call Botswana currently top-ranked in the region.
Why this is bounded
The checked CPI ranking puts Seychelles and Cabo Verde ahead, so a top-ranked frame would be stale.
Evidence status
FACT Economic framing
CONFIRMED / SCOPED
Corrected wording
Scope the Data Tracker to participating public-sector registries, not literally all data.
Why this is bounded
The transparency model is real, but the coverage boundary matters for remedy claims.
Evidence status
FACT Economic framing
STALE
Corrected wording
Treat the line as outdated unless a current ranking is sourced; Cambridge UK is ahead in the checked note.
Why this is bounded
The economic context should not rely on a dated city-branding statistic.
Evidence status
FACT Economic framing
NEEDS-PRECISION
Corrected wording
Cite those warnings only with disclosure that they are partly Google-funded.
Why this is bounded
Funding disclosure is required to preserve reader trust in comparative regulatory arguments.
Evidence status
FACT Economic framing
UNPROVEN
Corrected wording
Do not conflate procurement, export licensing, or revolving-door links with party donations.
Why this is bounded
The checked spyware file found no primary-source-documented European party donation case, including no Palantir or Thiel European party-donation record.
Evidence status
FACT Spyware / thesis-favorable rejected claim
NEEDS-PRECISION
Corrected wording
Name the three de-listed individuals and state that Tal Dilian and the corporate entities remain designated.
Why this is bounded
The corrected record is narrower than a generic sanctions-lifted line: Andrea Gambazzi, Sara Hamou, and Merom Harpaz were removed on 30 December 2025 after reconsideration petitions.
Evidence status
FACT Spyware
NOT FOUND
Corrected wording
Say they were not found in the EU Transparency Register export checked, not that absence proves they never lobby.
Why this is bounded
The checked register export found no NSO, Intellexa, Cytrox, or obvious corporate-variant entry; absence from the register is an absence finding, not proof of no lobbying.
Evidence status
FACT Spyware / thesis-favorable rejected claim
CONTESTED / NEEDS-PRECISION
Corrected wording
Treat the GDP line as a projection and the tourism line as a single-year estimate; note that the democracy-rating premise is contested.
Why this is bounded
This economic-framing material is not clean fact because Freedom House kept the United States in its Free category while V-Dem downgraded it.
Evidence status
FACT Economic framing
UNSOURCED
Corrected wording
Keep the Acus name, but do not use the Latin etymology gloss unless sourced.
Why this is bounded
The name is confirmed; the etymological interpretation is collaborator gloss.
Evidence status
FACT Sweden
Verification practice
Load-bearing claims are checked against primary records where possible, recorded in the verification log, then folded into the research files. Failed or overstated claims are not adopted; they are parked in the ledger.
Challenge route
To challenge a claim, cite the exact sentence, identify whether the issue is source support or analysis, and provide the primary record that corrects it. Successful challenges should update the research file, the verification log, and this public methods page.