Part 1: Purpose and scope
1.1 What TMI is
Modern Insurgent is an independent research and media organization that explains armed movements, insurgencies, political-violence actors, and the broader landscape of organized resistance. Our tagline, “from the front line, on the record,” describes the work precisely: we document who these actors are, what they want, what they do, and how states and societies respond, and we do so in a consistent, sourced, and non-partisan register.
We are not an advocacy organization, an intelligence service, or a wire agency. We do not take sides between the actors we cover, and we do not amplify any actor’s own messaging on its behalf. We aggregate, attribute, and contextualize.
1.2 What this Framework governs
This Framework governs every public output TMI produces. As of this version those outputs are:
the Database of Insurgency Reports; News and Analysis articles; the Watchlist feed of news briefs and its featured Key Players strip; Strategy Reports; Productions, comprising documentaries, podcasts, interviews, creative work, and features carried by other outlets; and the Interactive Atlas, which visualizes the Database.
It also governs AI-assisted production, including the Watchlist Aggregator described in Part 8. Any new content type introduced after this version must be brought under this Framework before it is published.
1.3 Why it is public
A framework that is enforced privately cannot be audited. Publishing ours allows readers, academics, and the actors we cover to see the rules we hold ourselves to, to judge our consistency, and to challenge us when we fall short. Transparency is a credibility instrument, not a courtesy.
Part 2: Editorial principles
Six principles bind all contributors. They are listed in priority order: where two principles appear to conflict in a specific case, the earlier one governs, and the contributor should raise the conflict with an editor rather than resolve it silently.
2.1 Report events, do not relay claims
The unit of TMI reporting is a verifiable event or a documented body of fact, not an actor’s own assertion about itself. A clash, a sanctions designation, a law, a protest, or a founding is reportable. An actor’s propaganda, manifesto, or recruitment material is evidence to be analyzed and attributed, never content to be passed through in the actor’s voice. We report on these actors; we do not redistribute their material.
2.2 Objectivity and non-partisanship
TMI describes; it does not endorse, condemn, or root. Contributors write as analysts, not advocates. The same descriptive standard applies to every actor regardless of ideology, methods, or sympathy. Where defenders and critics of an actor hold opposing factual or interpretive positions, we present the strongest honest version of each and attribute it, rather than adjudicating in TMI’s own voice.
2.3 Non-speculation
TMI states what is sourced and labels what is not. Inference is permitted when it is identified as inference and grounded in cited fact. Prediction, motive attribution without evidence, and unsourced assertion are not permitted in TMI’s voice. When the record is uncertain or contested, we say so plainly.
2.4 Sourced at all times
Every factual claim must be traceable to a source a reader can check. No brief, report, or article asserts anything in TMI’s voice that cannot be pointed to. The sourcing standard is set out in full in Part 6.
2.5 Neutral language
We describe violence; we do not use it to frame. Loaded or honorific terminology, slurs, and the actors’ own euphemisms are avoided in TMI’s voice and quoted with attribution where they are themselves the subject of analysis. The classification of an actor is an analytical output governed by Part 4, never a rhetorical device.
2.6 Do no avoidable harm
TMI analyzes political violence without functioning as an instruction manual for it, a targeting aid, or a vector for an actor’s reach. This principle governs Strategy Reports in particular (Part 7) and constrains operational detail across all content types.
Part 3: Definitions and assumptions
Stated definitions are what keep classification consistent across hundreds of entries and dozens of contributors. The following terms carry the meanings below throughout TMI’s work.
3.1 Insurgency
Our working definition is adapted from NATO Counterinsurgency (COIN) doctrine, which identifies three conditions: action by an organized group; a goal of political change directed against a ruling order; and the use of violence or subversive activity. Under TMI’s adapted definition an actor must satisfy the first two conditions, organization and a goal of political change. The third, violence or subversion, is treated as a possible characteristic rather than a prerequisite. Not all insurgencies are violent, and TMI deliberately covers non-violent movements that meet the first two conditions.
3.2 Violence
Violence is any action or behavior that has caused pain, death, or physical damage. The term is descriptive and applied to documented acts only. It is never used to characterize an actor in the abstract, only to record whether definite actions have hurt, killed, or damaged a person or thing.
3.3 Actor, group, and movement
An actor is any entity TMI tracks as the subject of a profile. A group is an actor with identifiable organization, membership, or command. A movement is a broader current of activity that may lack hierarchy, leadership, or formal membership and is held together by shared aims rather than structure. TMI profiles both groups and movements; the distinction is recorded in the analysis rather than imposed on it.
3.4 Event
An event is a discrete, datable occurrence involving a tracked actor that a reader could in principle verify from outside TMI: an attack, a clash, a protest, an official response, a designation, a founding, a split, or a dissolution. Events are the unit of the Watchlist and the trigger for updates to the Database.
3.5 Source tier
Every source TMI cites carries a tier recording its institutional character. The four tiers, official, established, OSINT, and affiliated, are defined and governed in Part 6. Tiers are shown to readers, never hidden.
Part 4: Group classification
4.1 The classification system
Each actor in the Database carries one of four color classifications. The classification is an analytical conclusion drawn in the Approach to Resistance section of the actor’s Insurgency Report (Part 5.1) and is the single field that drives the actor’s marker color on the Interactive Atlas.
Red: armed and/or violent.
Blue: non-violent, focused on political activism.
Green: ambiguously inactive, disappeared, or splintered.
Pink: hacktivist organizations using digital resistance.
4.2 Mapping to the data model
In the content management system the four colors are stored as the actor’s status value: Red as violent, Blue as nonviolent, Green as inactive, and Pink as hacktivist. A separate size value (very_small, somewhat_small, somewhat_large, large) records the actor’s relative influence or scale and sets its Atlas marker size. These two values, together with the actor’s coordinates, are what allow a single Database record to drive both the article and its map marker.
4.3 Assignment and reclassification
Classification follows from the documented record, not from an actor’s self-description or from the gravity of its stated aims. An actor that espouses radical goals but has taken no violent action is classified by what it has done. Because actors change, classification is revisited whenever a tracked event materially alters the picture and, in any case, at the regular review interval set in Part 9. A change of classification is a substantive edit and is recorded as such.
Part 5: Content types and required structure
Each content type below has a fixed structure or a defined standard. Holding to a shared structure is a bias-control measure: it forces every actor and every story through the same questions in the same order, so that what differs between entries is the evidence, not the framing.
5.1 Insurgency Reports (the Database)
The Database is TMI’s core archive and the backbone of the site. Each entry is a scholarly profile of a single actor and follows a seven-part structure in fixed order:
- Overview: a plain introduction to the actor: what it is, where it operates, and why it is tracked. Section headings may name the subject (for example “Movement Overview”) but the function is fixed.
- History and Foundations: the relevant historical and founding context, including the conditions and events that produced the actor.
- Objectives and Ideology: the actor’s stated aims and ideology. This section carries extra sourcing scrutiny because it is where an actor’s self-description is most likely to enter the text; claims about motive and belief must be attributed, not asserted.
- Political and Military Capabilities: the actor’s regional influence, organizational reach, and, where applicable, militant capacity. Capability is described from evidence, not estimated speculatively.
- Approach to Resistance: how the actor pursues its aims. This section is where the Part 4 classification is reasoned and stated.
- Relations and Alliances: the actor’s regional dynamics: allies, rivals, sponsors, and adversaries.
- Works Cited: full source transparency. Writers may use their preferred citation style provided every factual claim in the body is traceable to a numbered or named entry.
Each report also carries the metadata that places it in the Database and on the Atlas: region, summary, author or authors, date, the status and size values from Part 4, coordinates, and a marker logo. A report may additionally be flagged for feature on the Watchlist Key Players strip.
5.2 News and Analysis
News and Analysis articles are long-form, by-lined analytical essays on a specific development, dynamic, or question. They are not bound to the seven-part structure but are held to the full principles of Part 2: every factual claim is sourced, analysis is separated from inference, and no actor is framed or favored. Articles carry a title, an optional subtitle or hook, one or more authors, a date, and a thumbnail, and link their bylines to the contributors’ team profiles.
5.3 Watchlist news briefs
The Watchlist is TMI’s running feed of short, dated dispatches on breaking developments involving tracked actors, paired on the page with a curated Key Players strip of featured Database entries. A brief is the smallest unit of TMI reporting and follows a fixed shape:
a headline that is terse and factual, with no clickbait; an event-type tag, one of [ATTACK], [CLASH], [PROTEST], [RESPONSE], or [OTHER]; the date and, where possible, the location; a body of roughly one to three sentences in neutral, wire-style prose; the related tracked actor or actors, linked to their Database profiles; and a sources block.
Briefs report events, never relay claims (Part 2.1). The same event must produce exactly one brief regardless of how many channels report it (the deduplication rule, Part 8.4). Briefs may be written by hand or produced by the Watchlist Aggregator under Part 8; both are held to identical standards, and the sources block follows Part 6 in either case.
5.4 Strategy Reports
Strategy Reports are analytical studies of the tactics and methods used in political violence, grouped into four standing categories: Assassinations and Targeted Killings; Explosives and Arms; Military Tactics and Operations; and Urban Warfare and Sabotage. They explain how a tactic functions in the world, why actors choose it, its strategic logic, and its countermeasures, drawing on the historical and scholarly record. Because this content is analytically sensitive, it is governed by a dedicated ethics standard in Part 7 in addition to the general principles. Each report carries a title, its category, a summary, an author, and a thumbnail.
5.5 Productions
Productions are TMI’s mixed-media work, organized into sub-types: documentaries; podcasts, across the organization’s series; interviews; creative productions; and features carried by other outlets. Productions are held to the same standards of objectivity, sourcing, and non-speculation as written work. Interviews are purely informational: they exist to record what a subject says, attributed to that subject, and are never structured to frame, endorse, or discredit the interviewee. The report-events-not-relay-claims principle (Part 2.1) applies: hosting a subject is not endorsing the subject, and an interview is presented as that subject’s account, not as TMI’s findings.
5.6 The Interactive Atlas
The Atlas is a world map of tracked actors. It is not a separate body of content but a visualization of the Database: each marker is derived from a Database record’s coordinates, classification color, and size. Because the Atlas is generated from Database records, it inherits their sourcing and classification automatically and carries no independent claims of its own. A small number of markers may link to News and Analysis pieces rather than to a profile; these are visually distinguished and are the only Atlas markers not backed by a full Insurgency Report.
Part 6: Sourcing and citation
6.1 The sourcing requirement
Every factual claim TMI publishes must be traceable to a source a reader can check. No output asserts anything in TMI’s voice that it cannot point to. This applies identically to a seven-part Insurgency Report and to a one-sentence Watchlist brief.
6.2 Source tiers
TMI does not draw only on established outlets, so each source is labeled with a tier that records its institutional character. The tier is shown to the reader, never hidden, because knowing what kind of source stands behind a claim is part of the claim. The four tiers, in descending order of independent authority, are:
Official or institutional: governments, intergovernmental organizations, courts, and sanctions registers.
Established media: wire services and recognized news outlets.
Independent OSINT: analysts, conflict monitors, and verified geolocators operating in the open-source space.
Group-affiliated: an actor’s own channels or those aligned with it.
6.3 Showing the sources
For events, the cluster of sources reporting an event is itself the citation: rather than citing a single source, TMI shows all sources covering one event side by side, each carrying its tier. This makes corroboration visible and turns deduplication into a feature rather than a cost, since “this event, covered by N sources, here they are” is both the citation and the confidence signal. Hand-written briefs and long-form articles may present their sources as a numbered or named list; the tiering principle still applies in spirit, and any claim resting solely on an affiliated source is identified as such.
6.4 Using affiliated material
Group-affiliated sources may be cited as evidence of what an actor says about itself, never as independent confirmation that what it says is true. An event sourced only to the affiliated tier is flagged as such, because the absence of independent corroboration is itself part of the story. TMI cites and analyzes such material; it does not republish or amplify it.
6.5 Verification and uncertainty
Contributors corroborate before asserting and prefer higher-tier and multiple-source confirmation for consequential claims, especially casualty figures, attributions of responsibility, and designations. Where the record is contested or thin, the uncertainty is stated in the text rather than smoothed over.
Part 7: Ethics standard for Strategy Reports
Strategy Reports examine assassinations, explosives, arms, and the tactics of armed struggle. The analytical value of this work and its potential for misuse sit close together, so the following standard governs it in addition to the general principles. An editor will hold or unpublish any Strategy Report that fails it.
7.1 Analytical, not instructional
A Strategy Report explains how a tactic functions as a strategic phenomenon: why actors choose it, what it achieves, what it costs them, and how it is countered. It does not function as a how-to. Reports describe tactics at the level of strategy, history, and consequence, not at the level of operational instruction.
7.2 No operational detail
Reports exclude content whose primary value would be to help someone carry out an attack: synthesis routes, device construction, procurement methods, defeat of specific protective measures, or step-by-step procedure. Where a source contains such detail, TMI draws the analytical point from it without reproducing the detail. The test is whether a passage informs understanding of a phenomenon or instead lowers the practical barrier to committing it; the latter is excluded.
7.3 Grounded in the established record
Strategy Reports are built on the historical, academic, and open-source record, using documented past cases to illustrate strategic logic. They do not present novel methods, and they do not solicit, host, or pass through material produced by actors to promote their own tactics.
7.4 Countermeasures and consequence in view
Because these reports treat tactics of harm, they situate each tactic in its consequences: its human cost, its effect on the actors who use it, and the countermeasures developed against it. This is both analytically complete and the posture that distinguishes study of a tactic from promotion of it.
Part 8: AI-assisted production and the Watchlist Aggregator
TMI uses AI assistance in production, most substantially in the Watchlist Aggregator, a system that watches a defined set of tracked actors, detects events meeting Watchlist criteria, and drafts briefs in the standard format. AI-assisted output is held to every standard in this Framework without exception, and to the additional rules below.
8.1 Disclosure
AI-assisted content is identifiable as such. A brief produced by the Aggregator records its provenance, including the fact of machine generation, in its metadata, and the public feed does not present machine-drafted text as hand-reported work.
8.2 Human accountability
No standard is relaxed because a machine drafted the text. A human-defined review gate stands between candidate output and publication: a brief is live only when it has cleared review and is marked published. Responsibility for every published brief rests with TMI, not with the tool.
8.3 Report events, not claims, at machine scale
The Aggregator’s unit of output is an event that meets a Watchlist tag, not an actor’s own post. A propaganda video is not, by itself, an event. The same report-do-not-relay principle that governs human briefs (Part 2.1) is the Aggregator’s core constraint and its compliance posture.
8.4 One event, one brief
Each real-world event appears exactly once. The failure mode of an automated feed is not sensationalism but the same event surfacing repeatedly because many channels posted it, which reads as noise and erodes trust. All reports of one event are clustered to a single brief keyed to a stable cluster identifier, and the count of clustered sources becomes the corroboration shown to the reader.
8.5 Tiered sourcing and confidence
Every Aggregator brief carries the tiered source block of Part 6. Confidence is a function of corroboration count and source tiers and is recorded with the brief. A brief resting only on affiliated sources is flagged, never silently published as settled fact.
8.6 Build order and the bounded promise
The Aggregator is built lane by lane. The policy and response lane, drawing on authoritative and easily cited sources such as sanctions registers, official statements, and established wires, is proven first. The combat and ground lane, drawing on OSINT and harder-to-verify channels, is enabled only once the pipeline is trusted on the first lane. TMI’s promise is completeness within its tracked actors and defined tags, not coverage of everything everywhere; that bounded promise is one TMI can keep, so gaps within it are accountable rather than excusable.
Part 9: Procedure, corrections, and governance
9.1 Editorial workflow
TMI’s standard writing procedure runs in five stages: topic selection; a first draft, typically allowing about two weeks including research; editor feedback; a second draft; and publication. Watchlist briefs run on a compressed version of the same path, with the review gate of Part 8.2 standing in for the draft cycle on machine-assisted items.
9.2 Updates
Insurgency Reports are reviewed and updated on a regular cycle, by default every six months, so that profiles, capabilities, and classifications stay current. A tracked event that materially changes an actor’s situation prompts an update ahead of the scheduled cycle. Updates are substantive edits and are recorded as such.
9.3 Corrections
When TMI publishes an error of fact, it corrects the record rather than quietly amending it. A correction notes what was wrong and what was changed. Because our credibility rests on showing our sources, it equally rests on owning our mistakes in the open. Readers who identify an error can reach the editorial team using the contact below.
9.4 Independence and conflicts of interest
TMI does not take sides between the actors it covers and accepts no arrangement that would condition its coverage. Contributors disclose any personal, financial, or organizational interest bearing on a subject they cover, and an editor reassigns or supervises the work where a conflict could compromise objectivity.
9.5 Funding
Reader donations are pooled in a dedicated account for future projects and are never used for personal purposes. Any use of funds requires the agreement of the full team. Funding sources do not influence editorial judgment, and no funder receives coverage consideration.
9.6 Governance of this Framework
This Framework is the standard against which TMI’s work is judged, including by readers. It is reviewed periodically and amended as the organization’s content evolves; material amendments are versioned and dated. Any new content type or production method must be brought under this Framework before it is published.
Contact
If we have omitted something you consider important, or if you have identified an error in our work, contact the editorial team at team@moderninsurgent.org.
Modern Insurgent · From the front line, on the record.