CAMO — assessing the reliability of information
Four questions, four levels of establishment, five rules. This framework is published so that our conclusions can be challenged using our own criteria.
The problem
The volume of available information has never been higher, and its reliability has never been harder to establish. Three phenomena are converging.
Circulation erases the trail. A piece of information repeated a hundred times appears confirmed a hundred times. It is confirmed only once — and often poorly.
The production of plausibility has been industrialised. Artificial intelligence systems produce, at zero cost, texts that are coherent, plausible and sourceless.
Decisions are made faster than verifications. The gap is filled by trust, habit or intuition.
In environments where information is costly to verify, that gap becomes a decision risk.
What this framework does, and does not do
CAMO answers a single question, and refuses to address any other:
The framework does not establish what is true. It establishes what is demonstrable given the sources available — and, just as importantly, what is not.
An instrument, not a general method
CAMO serves one precise operation: establishing facts, where the material under examination is a piece of information or a counterparty.
It says nothing about structuring an organisation, ranking strategic options or steering a decision. Those tasks belong to other instruments.
When to apply it
CAMO applies to any assertion meeting the following three conditions:
- It carries a consequence. If it is false, a decision changes
- It is not directly observable by the recipient of the analysis
- It is capable of being established through accessible sources
An assertion that fails the third condition falls outside CAMO: it belongs to the realm of hypothesis, and must be treated as such.
Not every assertion deserves the same effort. The intensity of the examination is set by the consequence of being wrong, never by the ease of verification.
An assertion that is easy to verify but carries no consequence can be set aside. An assertion that is hard to verify but decisive must be addressed — and if it cannot be, that must be stated in writing.
The four dimensions
They are examined together. None is sufficient on its own, and the order of examination does not matter.
Information never exists independently of the conditions under which it was produced. Understanding those conditions does not discredit the information: it allows its weight to be properly assessed.
- Who produced the information, and in what capacity
- For what purpose: regulatory filing, corporate communication, legal obligation, court decision, journalistic work, private exchange
- Did the producer have direct access to the facts being reported
- Is there a duty of truthfulness, and a sanction for breaching it
- Who benefits from its circulation, and in what capacity
- Produced spontaneously, or in response to a request
A client reference mentioned during a negotiation is produced in a context where completeness is not expected. The absence of any mention of a dispute is not a lie: it is a predictable consequence of the context.
Age does not invalidate information. It determines its present-day reach. Some information never ages; some expires within weeks.
- Date of production, as distinct from the date of publication or last update
- Reasonable validity period for this type of information
- Events since then that may have altered it
- Any revision, retraction or update
- Is the source under any obligation to keep it current
A frequent pitfall. The date displayed on a web page is almost always that of its last technical modification, not that of the information’s production. The two must never be confused.
This is the most frequently misapplied dimension. Repetition is not confirmation.
- Number of sources reporting the information
- Genuine independence: do they cite one another, trace back to a common origin, belong to the same group
- Nature of the convergence: agreement on the facts, or reuse of the same wording
- Existence of diverging sources, and any explanation for the divergence
- Is the absence of contradiction meaningful, or does it reflect a lack of attention
The decisive dimension. Information whose origin cannot be traced can never be rated as established, however many times it has been repeated.
- Can the chain of transmission be traced back to its point of origin
- Is the primary source accessible, consultable, and capable of being relied upon against third parties
- Nature of the source: authentic instrument, registry entry, court decision, official publication, statement by a party, testimony
- Broken chain, inaccessible link, uncertain attribution
- Alteration in transmission: rounded, generalised, taken outside its original scope
A figure presented as industry-wide but drawn from a study covering a single country has been altered in transmission. Its source is identifiable; its scope has changed.
The four levels
A verdict applies to an assertion, never to a person or an organisation. This distinction is fundamental: we do not rate an entity, we rate what can be asserted about it, at a given date, on the basis of specified sources.
The assertion rests on an accessible primary source, or on concordant independent sources. It can be relied upon against third parties and carried into a decision without particular reservation.
Cumulative conditions. Origin: primary source identified and accessible. Multiplicity: two concordant independent sources, or one authoritative primary source. Age: information current, or whose age has no bearing on what is asserted. Context: identified and not undermining its weight. No contradictory element found.
The assertion is supported by converging elements, without the chain of evidence being complete. It can ground a decision, subject to an express reservation.
Conditions. A single dimension shows a weakness, and that weakness is identified. Typical situations: primary source exists but is not accessible within the timeframe; credible and converging secondary sources without access to the origin; information previously established whose current validity could not be confirmed; non-neutral context with no contradictory element.
The available elements make it possible neither to support nor to dismiss the assertion. A decision resting on it carries a risk that must be explicitly accepted.
Conditions. At least one of the following: sources exist but are dependent, or trace back to a single unverifiable origin; contradictory elements not resolved; a context making the information non-neutral, with no possibility of cross-checking; an age that makes present-day validity impossible to assess.
No accessible element supports the assertion.
Conditions. No accessible source identified; or an assertion resting solely on the statement of the interested party; or a break in the chain of transmission preventing any trace-back.
This verdict indicates that the assertion could not be established with the sources available, within the scope and timeframe of the examination. It is neither an accusation, nor a denial, nor a value judgement.
A significant share of what is true cannot be established from open sources. Mistaking the absence of evidence for evidence of absence is the gravest error an analyst can make — and the most frequent.
"Not established" has no colour of its own: it is depicted by the absence of fill. A red background would read as a denial, which this rule prohibits.
The pattern — solid, half, hatched, empty — carries the information independently of colour. The marker remains legible in monochrome, in print, and to a colour-blind reader.
The five rules
Five, no more. A framework with twelve rules has none: nobody remembers them, so nobody applies them.
Each one binds every user of the framework. Each is constitutive — remove it and the framework breaks. Each blocks a real and frequent failure of analysis.
Prohibited: concluding that an entity is reliable or dubious. An accumulation of "not established" verdicts produces no overall judgement — it produces a list of what remains to be established.
Without this rule, the scale becomes an entity-rating system. The framework changes in nature, and its user is exposed to legal risk.
Prohibited: treating a "not established" verdict as a negative signal, an implicit reservation or a ground for rejection.
This is the most frequent and the most damaging error. It turns an instrument of analysis into a machine for exclusion.
Required: before any count, check whether the sources cite one another, trace back to a common origin or belong to the same group. The count of independent sources appears separately from the count of sources consulted.
Without this rule, the Multiplicity dimension becomes a tally, and the framework mechanically validates whatever circulates widely.
Required: every verdict carries its date, its scope and its source count. Reusing it in another context, or at another date, requires re-examination.
Prohibited: citing a verdict without its bounds. A verdict detached from its conditions is no longer a verdict — it is an opinion equipped with borrowed vocabulary.
Required: every analysis includes an explicit list of what could not be established, and distinguishes what fell outside the scope from what was sought without result.
Prohibited: delivering a clean conclusion by absorbing the doubt. A conclusion without stated limits is not more useful — it is merely more comfortable.
This is the rule that separates a defensible decision from a confident one. It is the purpose of the framework.
1. A verdict applies to an assertion, never to a person.
2. "Not established" does not mean "false".
3. Repetition is not confirmation.
4. A verdict holds at a date, within a scope, with sources.
5. What could not be established is stated in writing.
The verdict sheet
Every deliverable ends with this sheet. Two of its lines appear in no competing report: the count of independent sources, distinct from the total count, and the section devoted to what could not be established.
- The exact identity between the registered beneficiary and the bidding organisation
- The actual amount and scope of the programme
- The quality of execution — no evaluation report published
An illustrative example, constructed to explain the sheet.
A complete example
Context. An international organisation is considering entrusting the implementation of a programme to a local organisation. That organisation states that it implemented a comparable programme, funded by a multilateral donor, completed two years earlier.
Assertion assessed. "This organisation implemented a programme funded by [donor] and completed in 2024."
Examination
Context. The assertion is produced by the organisation itself, as part of a bid. A non-neutral context: completeness is not expected there, and the selection of favourable elements is predictable. This does not discredit the assertion, but it forbids stopping at it.
Age. The programme has been completed for two years. The assertion concerns a past fact: its age has no bearing on its veracity. It does, however, bear on what can be inferred about current capabilities — the team that ran the programme may no longer be in place. That inference belongs to another assertion, to be assessed separately.
Multiplicity. Three mentions of the programme were identified online. On checking, two reproduce the organisation’s own press release. The third comes from a sector directory that does not indicate its source. Apparent multiplicity: three. Actual multiplicity: one.
Origin. The donor’s transparency portal records a grant awarded over the stated period, in the country concerned, on a matching theme. The beneficiary appears there under a similar but not identical name. No contractual document is published.
Verdict
Probable. The origin is partially established — a matching grant exists and appears in a high-quality primary source. The exact match of the beneficiary could not be confirmed. No contradictory element was found.
What could not be established: the exact identity between the registered beneficiary and the bidding organisation; the actual amount and scope of the programme; the quality of its execution; whether the people who ran it are still in post.
The assertion can be retained subject to reservation. A single question put to the donor, or the production of the contract by the organisation itself, would move this verdict to "Established" within days. The cost is zero; the reduction in uncertainty is substantial.
The verdict is neither a rejection nor a validation. It is a map of what is known and what is not, together with an indication of the shortest path to resolving what remains.
Limits and conduct
What the framework does not do
- It does not determine the truth. It determines what is demonstrable with the accessible sources, at a given date
- It does not measure importance. A piece of information can be perfectly established and irrelevant to the decision. Prioritisation belongs to the analysis
- It does not apply to intentions or to the future. What a party will do, or will be capable of tomorrow, cannot be established
- It replaces no authority. Neither a court, nor a forensic expert, nor a statutory auditor, nor a regulator
- It is not an instrument of accusation. A "not established" or "uncertain" verdict authorises no imputation. It authorises caution
Principles of conduct
- A verdict is dated. It holds at the date of its issue, with the sources of that time
- A verdict applies to an assertion, never to a person
- What could not be established is stated in writing
- The scope is declared. The absence of examination is never presented as an absence of risk
- Doubt is communicated, never absorbed
Automated systems
The output of a language model is information without an origin. By construction, it can never exceed the verdict "not established" — unless the assertions it contains are re-examined one by one against the four dimensions.
This rule applies to SAIME3i’s own work. We use these systems in production; none of their outputs enters a deliverable without having been put through the framework.
We apply this framework through a deterministic validation architecture, upstream of any conclusion assisted by an automated system. The establishment of facts and the drafting of conclusions are separate steps. Third-party data is anonymised at every step involving an external system. Every verdict carries a control reference.
The framework is public so that you can challenge our conclusions. Our implementation is our production tool: it is auditable by your compliance function, it is not published.
Licence and citation
This framework is published under a Creative Commons Attribution — NoDerivatives 4.0 licence. No registration is required, no permission needs to be sought.
- Use it in your own work
- Reproduce and distribute it
- Use it commercially
- Teach it and cite it
- Apply it to challenge our conclusions
- You credit the author and the source
- You do not modify it
- You state the version used
- You do not present an altered version as being CAMO
Recommended citation
Saleh, H. (2026). CAMO 2.0 — A framework for assessing the reliability of information. SAIME3i. https://www.saime3i.com/camo
This page is printable. Use your browser’s print function to obtain a document version.
Version history
| Version | Date | Change |
|---|---|---|
| 1.0 | 2025 | First formalisation. Three-level verdict: reliable, suspicious, uncertain |
| 1.2 | 2026 | Deterministic validation layer added. Internal use |
| 2.0 | August 2026 | Scale redesigned into four levels rating the assertion, not the entity. Framework published. |
The change of scale introduced in version 2.0 is substantial: earlier verdicts cannot be automatically converted into the new grid.
Any future change will result in a numbered, dated version, published on this page. Analyses already produced remain governed by the version applicable at their date of issue.
Challenging this framework
If you believe a rule is poorly worded, a definition is confusing or a use case is not covered, write to contact@saime3i.com. Reasoned objections are examined and, where well-founded, incorporated into the next version.
You can use it on your own. Or ask us to apply it.
This framework is designed to be used without us. If you would rather entrust a verification, a diagnostic or a market entry, the scoping session takes thirty minutes and commits you to nothing.
You leave with a written note that belongs to you, whether we work together or not.
Discuss a situation