Evidence Before Position: Methodology, Privacy Boundaries, and Terms of Engagement

Project charter

International Intelligence is an independent, information-only public-intelligence research publication. Its purpose is to explain public power through records: authorities, budgets, contracts, policies, audits, litigation, disclosures, and carefully attributed reporting. It is not an intelligence tasking service, a secure drop, a legal practice, a crisis channel, a hacking guide, or a platform for targeting private people.

The project’s governing idea is evidence before position. Political alignment, institutional prestige, urgency, and narrative appeal do not determine whether a claim is true. The source trail does. Strong evidence receives more weight; contrary records are not hidden; uncertainty is stated; and corrections become part of the public memory of the work.

1. Proof before heat

Every material claim should be written so that another reader can inspect the evidence and understand the limits. A research note should identify the exact source, date, source class, preservation status, relevant excerpt or field, transformations performed, known gaps, and the evidence that would alter the finding. A link alone is not a source note: pages change, URLs break, documents are replaced, and context can disappear.

A defensible finding states what the record proves, what it does not prove, and what would force a revision.

The project distinguishes a source’s identity from a source’s evidentiary function. A respected organization may publish an explanatory article that links to a court order. The article is reporting or analysis; the order is a court filing. A government webpage may summarize a contract; the signed contract is the primary source. Classification follows the item actually relied upon.

2. Evidence Ladder

3. Confidence labels

Evidence class and confidence answer different questions. The tier describes what kind of source supports a claim. The confidence label describes the project’s present judgment after considering authenticity, independence, consistency, recency, completeness, and contradiction.

• Confirmed: directly established by authentic primary or official evidence, with no material contradiction identified.
• Corroborated: supported by multiple independent sources or reporting matched to an underlying record.
• Likely: evidence points strongly in one direction, but a decisive record is missing.
• Inferred: a reasoned conclusion from visible records that do not state the conclusion directly.
• Disputed: a material, sourced disagreement exists and is presented.
• Stale: once supported, but the relevant system, policy, organization, or record may have changed.
• Unknown: the available record does not answer the question.
• Unverified: not supported strongly enough for publication as fact.

A confidence label is not permanent. New records can strengthen, narrow, dispute, or overturn a finding. The date of the judgment is therefore part of the finding.

4. Follow the data flow

High-impact systems are often difficult to audit directly. Their code may be proprietary, their model may change, and affected people may not know which system influenced a decision. The project therefore begins with the institutional chain around the technology.

1. Collection point: What data enters the system, from which devices, forms, brokers, platforms, sensors, or agencies?
2. Authority line: Which law, policy, vote, grant, contract, memorandum, or official approved the collection or use?
3. Vendor and data flow: Who can query, export, enrich, retain, share, or delete the data? Which subcontractors or cloud services participate?
4. Decision point: What output affects a person—match, score, alert, ranking, denial, investigation, moderation action, or referral?
5. Human review: Does a reviewer have the information, time, authority, and obligation to challenge the automated output?
6. Audit gap: Which access logs, purpose codes, validation results, misuse discipline, correction rules, or deletion controls are absent?

Procurement records are especially useful because they can reveal scope, price, accuracy claims, data ownership, retention, integration, maintenance, and termination terms without requiring unauthorized access to a deployed system.

5. Authorization is a bright line

Public-interest purpose does not create consent. International Intelligence does not authorize or encourage bypassing access controls, probing systems without permission, evading rate limits, impersonating people, obtaining credentials, scraping protected areas, trading exploits, distributing malware, or acquiring unlawfully obtained private data. Defensive research belongs within explicit authorization, established testing programs, and responsible-disclosure procedures.

The records toolkit is designed for lawful requests to public bodies and other institutions. Users remain responsible for applicable law, deadlines, exemptions, fees, appeal rules, protective orders, contractual duties, and jurisdiction-specific advice. The site provides information and drafting assistance, not legal representation.

6. No doxxing, harassment, or target dossiers

Accountability focuses on systems, institutions, vendors, public decisions, and material professional conduct. The mere availability of personal information does not make publication necessary. The project rejects use of its sources or templates to expose credentials, home addresses, medical details, private beliefs, family members, minors, victims, or unrelated bystanders; facilitate stalking or harassment; or build targeting packages about private people.

Named individuals may be relevant when they exercise public authority, sign a contract, submit sworn testimony, direct a program, lobby for a policy, or hold a material corporate role. Even then, publication should be proportional to the public-interest claim. Personal detail that does not prove the claim should be removed.

7. Minimize before publication

Data minimization is an editorial act, not merely a privacy notice. Before publication, reviewers ask whether each personal field is necessary to establish the public claim. Redaction or omission should ordinarily cover private residential addresses, personal telephone numbers, personal email addresses, account credentials, precise location unrelated to the claim, medical or educational detail, identifiers of minors and victims, and metadata that exposes uninvolved people.

Minimization also applies to retention. Raw source files that contain sensitive material should be access-limited and retained only as long as the research, legal, historical, or correction need justifies. Public normalized feeds should carry only fields necessary for discovery and verification.

8. Source protection without false promises

Ordinary web forms and email are not secure channels for classified material, credentials, imminent-danger reports, or highly sensitive source communication. This release intentionally includes no upload form or anonymous drop. Contact addresses are for ordinary editorial, correction, and security correspondence. The project must never imply technical protections it has not independently designed, tested, documented, and maintained.

When source-sensitive reporting is contemplated, the method of communication, risk model, identity exposure, provider access, device security, metadata, retention, and legal process should be evaluated before material is transmitted. A warning is more ethical than a false promise.

9. Artificial intelligence as an assistant, never an authority

Automated tools may help sort records, detect likely duplicates, suggest topics, compare tables, draft translations, or summarize material for internal review. They may introduce omissions, fabricated relationships, mistranslations, false confidence, and hidden bias. Their output is therefore a work product to verify, not evidence to cite.

Every published material claim must be traceable to a human-inspectable source. A reviewer owns the final wording, evidence tier, confidence label, redactions, and translation. Where a machine-assisted translation has not received competent review, the site should say so or refrain from publishing it as a definitive edition.

10. Corrections and right of reply

A serious archive treats correction as evidence of method, not embarrassment. Correction requests should identify the affected page, exact disputed language, supporting record, requested change, and the requester’s relationship to the matter. Reviewers should record the issue, evidence received, decision, confidence change, visible edit, date, and any practical steps taken to repair downstream copies.

A materially named person or organization may submit a response that confirms, disputes, narrows, or adds context. A right of reply is not a veto. Unsupported denial does not outweigh stronger records, but it should not be concealed when it materially affects how the evidence is understood.

Privacy errors receive priority. If a page unnecessarily exposes a bystander, victim, minor, credential, or sensitive private detail, the information should be removed promptly while the project determines what public correction note is appropriate and whether cached or derivative copies can reasonably be repaired.

11. Attribution, copyright, and source terms

Each external source remains subject to its own terms, copyright, license, authentication requirements, and attribution rules. A public API is not a blanket license to republish every field for every purpose. Connectors should collect the minimum needed for public-interest discovery, link to the canonical source, identify transformations, and avoid implying endorsement.

International Intelligence’s original writing and code are provided under the notices included in the deployment package. Government records may have different status; third-party reporting and analysis remain the property of their publishers. Quotations should be limited and attributed. Bulk mirroring, commercial model training, or redistribution outside the project’s documented purpose requires an independent rights review.

12. Neutrality and false balance

Neutrality means applying the same evidentiary method to allies, opponents, agencies, companies, and advocacy groups. It does not require treating an unsupported denial as equal to a signed record or manufacturing a counterclaim where none exists. The project attributes opinions, distinguishes advocacy from evidence, and states conflicts and limitations.

Editorial conclusions may be firm when the record is firm. The discipline is not to avoid judgment; it is to make the route to judgment inspectable and correctable.

Terms summary

This charter is an editorial and operational standard, not legal advice. Questions involving active litigation, statutory deadlines, classified information, personal safety, or jurisdiction-specific obligations require appropriately qualified assistance.