Evidence layer
Probative surfaces brought into scope by this page
This page does more than point to governance files. It is also anchored to surfaces that make observation, traceability, fidelity, and audit more reconstructible. Their order below makes the minimal evidence chain explicit.
- 01Evidence artifactfalse-neighbors.json
- 02Evidence artifactsemantic-proximity-separation.json
- 03Evidence artifactcommon-misinterpretations.json
false-neighbors.json
/false-neighbors.json
Published surface that contributes to making an evidence chain more reconstructible.
- Makes provable
- Part of the observation, trace, audit, or fidelity chain.
- Does not prove
- Neither total proof, obedience guarantee, nor implicit certification.
- Use when
- When a page needs to make its evidence regime explicit.
semantic-proximity-separation.json
/semantic-proximity-separation.json
Published surface that contributes to making an evidence chain more reconstructible.
- Makes provable
- Part of the observation, trace, audit, or fidelity chain.
- Does not prove
- Neither total proof, obedience guarantee, nor implicit certification.
- Use when
- When a page needs to make its evidence regime explicit.
common-misinterpretations.json
/common-misinterpretations.json
Published surface that contributes to making an evidence chain more reconstructible.
- Makes provable
- Part of the observation, trace, audit, or fidelity chain.
- Does not prove
- Neither total proof, obedience guarantee, nor implicit certification.
- Use when
- When a page needs to make its evidence regime explicit.
Advertising brand safety vs brand safety in AI answers
Advertising brand safety aims, among other things, to prevent ads from appearing beside violent, illegal, hateful, deceptive or incompatible content. The risky content sits beside the advertisement.
In an AI answer, the brand may be integrated into the content itself: the system produces a sentence, comparison, attribution or recommendation. The risk is semantic and synthetic.
| Advertising | AI answer |
|---|---|
| Placement and adjacency controls | Limited control over external synthesis |
| Brand remains distinct from neighbouring content | Brand may be subject, object or example of the claim |
| Risk measured by context categories | Risk measured by claim, attribution, time and decision |
| Inventory can be excluded | Correction depends on sources, product and re-observation |
Example
An ad for a manufacturer appears beside an unrelated product-recall article: adjacency risk. An AI answer states that the manufacturer itself was recalled: answer risk. The interventions differ.
Legitimate criticism vs faulty association
A correctly attributed adverse decision is not an unsafe environment to erase. Assigning it to the wrong namesake, removing the date or turning an allegation into fact is an answer-level brand safety failure.
Escalation rule
Classify the claim first: accurate, disputed, attributed, outdated, confused or unverifiable. Then assess severity and decision proximity. An advertising policy cannot simply be copied onto an answer system.