Fact-Checking Policy

Purpose

Fact-checking at Cyber Insight EU is the structured process of examining whether published wording is supported by reliable evidence, accurately attributed, expressed in proportion to certainty and consistent across the headline, body, images, captions, metadata and disclosures. Fact-checking is not a decorative label. Cyber Insight EU does not describe an article as fact-checked unless meaningful verification has occurred. The author remains responsible for the accuracy of submitted work, and editorial review adds an independent layer of scrutiny rather than transferring all responsibility away from the author.

Proportionate Verification

Not every statement requires the same level of review. Verification intensity depends on the significance of the claim, potential harm, public interest, source reliability, degree of dispute, whether the subject is a private person, whether wrongdoing is alleged, whether the information concerns health, finance, elections or safety, whether anonymity is used, whether material may be synthetic or manipulated, and whether the claim may create serious legal or reputational consequences. High-risk claims require stronger evidence and more detailed records than ordinary background information.

Author and Editor Responsibilities

Before submission, the author identifies evidence supporting material claims, retains original documents and notes, distinguishes direct knowledge from secondhand information, checks names, dates, places and titles, verifies quotations, discloses uncertainty, identifies interested-party sources, reveals material conflicts, declares material automated-tool use, preserves calculations and alerts the editor to legal, privacy or safety concerns. Difficulty obtaining evidence does not justify unsupported certainty.

The editor independently tests claims in proportion to importance. This may include reviewing primary documents, checking recordings, opening cited sources, contacting institutions, comparing independent sources, reproducing calculations, examining images and metadata, checking geographic and chronological consistency, seeking expert explanation, contacting affected parties, reviewing legal terminology, checking conflicts and recording unresolved uncertainty. A claim is not accepted merely because the author is experienced, well known or professionally qualified.

Claim Classification

Cyber Insight EU distinguishes a verified fact from an allegation, estimate, inference, prediction, opinion, quotation and official finding. A verified fact is supported by reliable evidence within the limits of that evidence. An allegation is a claim not finally established. An estimate is an approximation whose method and uncertainty are explained. An inference is a conclusion drawn from established facts and identified as such. A prediction concerns a future outcome and states its assumptions. Opinion is a judgment or argument whose factual premises remain checkable. A quotation proves that a person made the statement, not that the statement is true. An official finding is described according to its authority, jurisdiction, stage and appeal status.

Source Evaluation

Sources are evaluated according to identity, expertise, proximity to the event, basis of knowledge, motivation, interest, independence, consistency, documentary support, reliability history and ability to be corroborated. An official source may be authoritative about what an institution decided but not necessarily about every factual assertion behind the decision. A company may be authoritative about its own announced policy but remains an interested party concerning performance, safety or social impact.

Cyber Insight EU prefers primary sources where available, including legislation, judgments, official decisions, public filings, original datasets, audited records, direct interviews, recorded speeches, original research, technical documentation and direct observation. Secondary reporting provides context, leads and interpretation. Where another publication produced the original reporting, Cyber Insight EU credits the originating work rather than implying independent discovery.

Quotations, Translation and Identity

A quotation is checked against audio or video, an original transcript, contemporaneous notes, an official written statement, a reliable published source or direct confirmation. It may be shortened when omission does not alter meaning. Words are not rearranged to create a more dramatic statement. Where audio is unclear, the uncertainty is resolved or the material is paraphrased.

Translated material preserves meaning, tone and cultural context. The editor considers whether qualified human review is required, whether legal or technical terminology carries special meaning, whether the quotation is disputed and whether automated translation was used. Material translation limitations are disclosed. Names are checked for spelling, order, diacritics, title, current or former role, organization, pronouns where relevant, legal identity and risk of confusion with another person.

Dates, Chronology and Location

Dates are checked against original records where possible. The editor verifies event date, announcement date, publication date, effective date, local time zone, sequence of events, whether a source was available at the relevant time and whether an update changed the timeline. Chronological inconsistency may reveal mistaken identity, manipulated media or unsupported causation. Location claims are examined through records, visual evidence, local knowledge or other reliable methods where material.

Numbers and Calculations

Numerical claims are checked for source, unit, currency, period, numerator, denominator, sample size, method, inflation, exchange rate, tax treatment where relevant, rounding, percentage-point versus percentage change, total versus average, revenue versus profit, downloads versus active users, registered accounts versus verified people and correlation versus causation. Calculations are reproducible from the stated method. Where a result depends on assumptions, the assumptions are visible.

Statistics, Research and Polling

Research coverage identifies study authors, institution, funding, publication status, peer-review status, design, sample, population, geography, period, uncertainty, limitations, competing evidence and whether the result has been replicated. A preprint is not described as established peer-reviewed evidence. Statistical significance does not automatically mean practical importance, and a study concerning a narrow population is not presented as universal.

Survey and polling coverage identifies the commissioner, conducting organization, field dates, sampling method, target population, sample size, question wording, response options, weighting, margin of error where applicable, non-response, geography and relevant limitations. A self-selected online poll is not presented as representative of the general population.

Company, Product and Market Claims

Company claims may concern user numbers, downloads, revenue, market position, performance, security, environmental impact, artificial intelligence capability, partnerships, certifications, rankings or regulatory status. Cyber Insight EU identifies whether the claim is independently audited, officially certified, internally measured, self-reported, projected, based on supplied documents or unverifiable from public evidence. Marketing language such as leading, revolutionary, most trusted, secure, green, largest, fastest or first is not adopted as fact without adequate support.

Legal Claims

Legal reporting identifies the jurisdiction, legal authority, procedural stage, whether the matter is civil, criminal, regulatory or administrative, whether an order is interim or final, whether an appeal exists and what the decision actually establishes. An allegation in a lawsuit is not a judicial finding, a charge is not a conviction, and an investigation is not proof of wrongdoing. Qualified legal review is obtained where interpretation is complex or consequential.

Images, Screenshots, Video and Social Media

Visual material is checked for original source, uploader, date, location, context, metadata, editing, synthetic generation, reverse-search results, weather, architecture, landscape, language, signage, shadows, lighting, audio consistency and prior publication. A genuine image may still be misleading if it is old, unrelated or falsely described. Detection tools may assist but are not treated as infallible proof.

Social posts are checked for account authenticity, date, deletion or editing, archive, context, impersonation, parody, translation, coordinated behavior, media authenticity and whether the account has direct knowledge. A verification badge does not establish the truth of a claim, and virality is not verification.

Anonymous Sources and Right of Reply

Anonymous-source claims receive additional scrutiny. The responsible editor knows the source’s identity, basis of knowledge, reason for anonymity, motivation, possible conflict, risk of retaliation, corroborating evidence and what can safely be disclosed. A single anonymous source supports publication only in exceptional circumstances and does not justify unsupported opinion or personal attack.

Contacting a criticized person or organization is also part of verification. A response may reveal factual errors, missing records, competing chronology, legal developments, mistaken identity, contextual omissions, conflicts or admissions. A response is not automatically true but is relevant evidence and is checked rather than reproduced uncritically.

Artificial Intelligence Verification

AI output is not a source. Every AI-generated fact, citation, quotation, case, statute, biography, title, calculation, research summary, translation, image description and technical explanation is checked against reliable non-AI evidence. Authors and editors are alert to fabricated citations, invented institutions, conflation of people, outdated roles, false certainty, missing jurisdiction, synthetic quotations, numerical errors and confident but unsupported explanations.

High-Risk Claims and Headline Review

Enhanced verification applies to criminal allegations, corruption, fraud, abuse, harassment, medical danger, investment claims, election information, national security, threats, confidential documents, leaked databases, hacking, children, sexual material, private individuals, discrimination, suicide, public-safety warnings and accusations likely to cause serious reputational harm. High-risk work uses a claim-by-claim record or equivalent evidence trail.

The headline is checked separately. The editor asks whether every implication is supported, whether an allegation is presented as fact, whether scale is exaggerated, whether the named person is correctly identified, whether a decisive qualification is omitted, whether causation is implied, whether a quotation is misrepresented, whether urgency is false and whether commercial status is concealed. A correct article with a misleading headline remains a publication failure.

Metadata and Final Reconciliation

Fact-checking applies to the SEO title, meta description, slug, category, tags, image metadata, author information, publication date, update date, article type, social preview and structured data. Machine-readable information does not contradict the visible page.

Before publication, the editor confirms that pending claims were resolved or removed, quotations match the source, numbers are consistent, names and dates are correct, source links work, disclosures are complete, the headline is supported, classification is accurate, the final page matches the checked draft and no formatting process introduced new errors.

Uncertainty, Failed Verification and Corrections

Where evidence is incomplete, Cyber Insight EU uses accurate qualification, such as according to the organization, the available records indicate, the estimate is based on, the evidence does not establish, the claim could not be independently confirmed, the sample is limited, the result remains preliminary, the allegation is denied or no final decision has been issued. Qualification clarifies uncertainty and is not used mechanically to publish an otherwise unsupported accusation.

When a material claim cannot be verified, Cyber Insight EU removes it, qualifies it, seeks additional evidence, delays publication, narrows the scope, identifies it as an allegation where independently newsworthy or declines the article. A deadline does not convert an unverified claim into fact. New evidence after publication may require correction, clarification, update, editor’s note or retraction.

Contact

Specific factual concerns may be sent to corrections@cyberinsighteu.com. The request should identify the article, exact statement, reason for concern, proposed correction and supporting evidence.