OPTARUSRISK & INTELLIGENCE← Return to main site
Analytical standard

Methodology & AI transparency

How OPTARUS collects, evaluates and communicates open-source intelligence—and where analyst judgement remains decisive.

Effective 21 September 2026 · Pre-launch version

Purpose and scope

OPTARUS produces decision-support assessments for maritime, energy, insurance and regional-risk users. FA-72 is a recurring Red Sea Commercial Transit Risk Monitor focused on material changes in capability, intent, geography, target profile, commercial exposure and warning indicators.

The service is not an incident feed. New reporting is included when it changes—or could credibly change—the standing assessment, a customer's exposure or a relevant decision threshold.

Analytical cycle

1 · DirectionDefine the intelligence requirement, customer decision and indicators that could change the judgement.
2 · CollectionReview openly available primary reporting, official notices, reputable journalism, specialist research, imagery and relevant industry reporting.
3 · EvaluationTest source access, credibility, independence, recency, corroboration and possible motive or bias.
4 · AnalysisSeparate intent, capability, sustainment, geography and consequence; test alternatives and identify intelligence gaps.
5 · DisseminationPresent the decision dashboard, material changes, map, network consequences and forward-warning ledger.
6 · ReviewCarry unresolved indicators and previous judgements into the next numbered issue.

Evidence and attribution

Each assessment distinguishes between four different evidential states:

  • Confirmed fact: sufficiently supported by reliable evidence or authoritative confirmation.
  • Attributed reporting: a claim accurately attributed to its source but not independently confirmed by OPTARUS.
  • Analytical assessment: the analyst's reasoned judgement based on the available evidence.
  • Unresolved hypothesis: a plausible explanation or linkage retained for collection and testing, not presented as established fact.

Maps and network diagrams follow the same discipline: supported, assessed and unresolved relationships must be visibly different.

Confidence language

Confidence is not a probability of an event occurring. It expresses how strongly the available evidence supports the judgement.

HighStrong, relevant and substantially corroborated evidence with limited material conflict.
ModerateCredible evidence supports the judgement, but important gaps, source limitations or competing interpretations remain.
LowThe judgement is plausible but rests on limited, indirect, contested or weakly corroborated evidence.

AI-assisted workflow

OPTARUS assessments are analyst-led and AI-assisted. AI does not independently publish, set confidence or make the final analytical judgement.

Permitted support

  • Discovery and organisation of potentially relevant open sources.
  • Translation assistance, extraction, comparison and timeline building.
  • Structured hypothesis testing, consistency checks and drafting support.
  • Formatting, quality-control prompts and preparation of repeatable report components.

Human controls

  • The analyst defines the collection requirement and decides what is material.
  • The analyst checks sources, quotations, dates, attribution and evidential caveats.
  • The analyst determines confidence and distinguishes reporting from assessment.
  • The analyst reviews named-person allegations, legal risk and the final publication.

AI output can be incomplete, outdated or incorrect. It is treated as a research and production aid, not as evidence in its own right.

Pre-distribution controls

Before external release, assessments are checked for attribution, current-versus-historical confusion, unsupported allegations, sanctions-status accuracy, defamation risk, data minimisation, UK data-protection concerns, criminal-offence information, export-control or ECJU sensitivity, visual clarity and advice boundaries.

Named individuals are included only when their identification is necessary and proportionate to the intelligence requirement. Public availability alone does not remove privacy or accuracy obligations.

Corrections and uncertainty

Material errors should be corrected transparently in the affected publication or the next numbered assessment. A changed judgement is not automatically an earlier error: each issue records what changed in the evidence and whether confidence increased, decreased or remained unchanged.

Professional feedback may be directed to Stephen J. Taylor through LinkedIn during the pre-launch period. A dedicated business contact channel will be published before subscriptions open.