Open sources, sourced judgments.Opensources,sourcedjudgments.
XS1 Intelligence builds AI models and capabilities for open-source intelligence: collecting, verifying and analyzing publicly and commercially available information across the web, imagery, corporate records, public markets and blockchains, and turning it into sourced, confidence-rated assessments.
Four areas, one standard of evidence.Fourareas,onestandardofevidence.
Each area has its own models and methods. Every one of them ends in the same place: a judgment a reader can trace to its sources, with its confidence stated.
From collection to a confidence-rated assessment.Fromcollectiontoaconfidence-ratedassessment.
Every engagement runs the same four stages, in the profession's own terms. Nothing reaches a reader without its provenance, and no judgment leaves without its confidence.
01 · Against a question
Collection
A collection plan maps sources to the partner's requirements. Material is preserved as it is captured: timestamp, URL, a SHA-256 hash and a handling log, so its integrity can be shown later.
Requirements
Collection plan
Preservation
Chain of custody
02 · Before it is used
Verification
Each item is checked for provenance and authenticity: geolocation, chronolocation, earliest copy, signs of manipulation. Sources are graded for reliability separately from the credibility of what they say.
Provenance
Geolocation
Chronolocation
Source grading
03 · What it means
Analysis
Records are resolved into entities and linked into networks. Structured techniques such as analysis of competing hypotheses keep alternatives in view, and independent corroboration is counted once per origin.
Entity resolution
Link analysis
Competing hypotheses
Corroboration
04 · Confidence-rated
Assessment
Key judgments come first, in estimative language. Confidence is stated separately as high, moderate or low, every judgment cites its sources, gaps are listed, and an analyst signs off before release.
Key judgments
Estimative language
Analytic confidence
Analyst sign-off
Estimative language
Likelihood in words. Confidence on its own line.Likelihoodinwords.Confidenceonitsownline.
Assessments use a fixed yardstick, so ‘likely’ means the same thing every time. How sure we are of the evidence is a separate grade, never folded into the likelihood.
Likelihood yardstick · ICD 203
Pick a term to rewrite the judgment
55–80%
0%25%50%75%100%
Key judgment
IllustrativeFictional example
We assess that the harbor channel is likely to stay closed to deep-draft traffic through Thursday.
Change the term and the likelihood changes. The analytic confidence does not.
Analytic confidence
We have moderate confidence in this judgment.
Low
Moderate
High
Graded from source quality, corroboration and method. It does not move with the likelihood.
The method is informed by public standards for open-source investigation and analysis, including the Berkeley Protocol on Digital Open Source Investigations, ICD 203 analytic standards and ISO/IEC 27037 guidance on preserving digital evidence.
What a reader receives
A judgment, its confidence, and everything behind it.Ajudgment,itsconfidence,andeverythingbehindit.
A specimen in the standing format, with notes on each part. Every judgment cites its sources, every source carries its grade, and what we could not establish is said as plainly as what we could.
IllustrativeSpecimen assessment · fictional subject and sources
Hover or tap a citation to see its source
XS1 Intelligence · Assessment
Fictional subject
Harbor channel blocked by a grounded vessel
Issued
Wed 06:00
Sources current to
Tue 18:00
Prepared for
Specimen
↳Freshness is stated, not assumed: when it was issued and how current its sources are.
01Freshness is stated, not assumed: when it was issued and how current its sources are.
Bottom line
A cargo vessel is aground across the harbor channel.
We assess that the channel is likely to stay closed to deep-draft traffic through Thursday.
↳Bottom line up front: the main judgment first, in a word from the estimative yardstick.
02Bottom line up front: the main judgment first, in a word from the estimative yardstick.
Analytic confidence
Moderate
LowModerateHigh
Three independent sources agree. One early notice contradicts and has not been withdrawn.
↳Confidence in words, on its own line, with its basis. Never a percentage, never mixed into the likelihood.
03Confidence in words, on its own line, with its basis. Never a percentage, never mixed into the likelihood.
Key judgments
The grounding almost certainly happened between 08:40 and 09:20 on Tuesday. The video’s shadows and the vessel’s AIS track agree on the window.
The vessel lies across the channel rather than at anchor beside it, so deep-draft traffic cannot pass.
It is unlikely that the partner’s two scheduled arrivals berth before Friday, given the two judgments above.
↳Each key judgment uses the same yardstick and cites what supports it.
04Each key judgment uses the same yardstick and cites what supports it.
Competing hypotheses
Older footage, recirculated: rejected. The earliest copy dates from Tuesday, and the shadows fit Tuesday morning.
At anchor, not aground:less consistent with the evidence, but the port notice still says the channel is open. Kept open until the notice is updated.
↳Alternatives considered, and why each was rejected or kept open.
05Alternatives considered, and why each was rejected or kept open.
Information gaps
No source covers salvage plans or tug availability. Either would change the Thursday judgment.
[3]Commercial satellite imageTue 10:40 · Usually reliable · probably trueB2
[4]Port notice: channel openTue 09:30 · Usually reliable · doubtfulB4
Preserved on capture: URL, time, SHA-256 hash · custody log, 4 entries
↳Each source graded twice (Admiralty Code): how reliable the source is, A to F, and how credible this item is, 1 to 6.
07Each source graded twice (Admiralty Code): how reliable the source is, A to F, and how credible this item is, 1 to 6.
Release
Reviewed and released by· Wed 05:40
↳An analyst reviews every consequential judgment and signs it off before release.
08An analyst reviews every consequential judgment and signs it off before release.
Who we work with
For teams that have to stand behind a judgment.Forteamsthathavetostandbehindajudgment.
XS1 Intelligence does not sell a data feed or a packaged product. It takes on contracted work for organizations whose decisions depend on what open sources can show.
Compliance and financial-crime teams
Enhanced due diligence, beneficial-ownership tracing, sanctions-exposure screening and blockchain exposure checks.
Investigators and legal teams
Digital open-source investigation, asset tracing and preservation-ready evidence with a documented chain of custody.
Corporate risk and security teams
Open-source threat and risk intelligence on the markets, suppliers and places an organization depends on.
Investment research teams
Filings analysis and event detection on point-in-time data: research signals and decision support, not investment advice.
The work we take on
Contracted research and engineering
Models and pipelines built around a partner's question: verification tooling, multilingual entity resolution, filings and disclosure models, tracing heuristics.
Contracted investigations and analysis
Scoped investigations and recurring assessments, delivered as sourced, confidence-rated reports an analyst has signed off.
Research partnerships
Joint work on the open problems below, scoped with the partner from the first conversation.
Research focus
The open problems we work on.Theopenproblemsweworkon.
Finding more material is rarely the hard part. The hard part is knowing which records are the same thing, which reports are independent, and how sure to be.
XS1 does not publish its team, partners or research. Qualified partners receive more detail after an initial conversation.
01
Multilingual entity resolution
Deciding which records describe the same company, vessel or wallet cluster across registries, scripts, transliterations and missing fields.
Registries
Transliteration
02
Corroboration, not syndication
Telling independent confirmation apart from one report repeated many times, so volume never passes for agreement.
Origin tracing
Reposts
03
Calibrated confidence
Models whose estimative language means what it says: when an assessment reads 'likely', it should hold up about that often.
Calibration
ICD 203
04
Point-in-time reasoning
Ordering events correctly when sources report late, are revised or run on different clocks, without letting later knowledge leak backwards.
Revisions
No look-ahead
Responsible use
Lawful sources, defined purposes, stated confidence.Lawfulsources,definedpurposes,statedconfidence.
Open-source intelligence is only as trustworthy as the limits it keeps. These are stated to every prospective partner before an engagement begins.
We decline work intended for
Harassing, stalking or unlawfully profiling individuals
Evading sanctions or law enforcement
Trading on material non-public information
Defending your own networks and systems from attack is the work of XS1 Cyber.
Lawful, available information
We work only with information that is publicly or commercially available and lawfully obtained, and only for lawful, defined purposes.
Every client and use case reviewed
We review every prospective client and use case before any engagement, and scope the work to its stated purpose.
Privacy by design
Collection and models are designed for applicable privacy and data-protection law, with data minimization and purpose limitation.
No access to private accounts
We do not access private accounts or bypass access controls.
No people search
We do not offer lookups of private individuals. A person enters a case only when an analyst records them, for a reviewed, defined purpose.
Every output carries its sources and a stated level of confidence, and consequential judgments are reviewed by an analyst before release. Blockchain attribution and other analytic judgments are probabilistic, and we say so.
News
Latest from XS1 IntelligenceLatestfromXS1Intelligence
XS1 Intelligence, XS1's open-source intelligence division, works across four capability areas: collection and verification, entities and networks, markets intelligence, and blockchain intelligence. One analytic method runs through all four and ends in sourced, confidence-rated assessments.
