Trust Centre Operations
Claims Standard
The public language standard for AI-assisted civic content.
Source-Controlled Doc
This public page is rendered from product-development/trust-and-safety/public-operations/claims-standard.md.
Purpose
The claims standard keeps Parliament People from overstating what AI-assisted content can safely promise.
Allowed Claims
- Source-grounded.
- Attributable.
- Reviewable.
- Transparent.
- Corrected when needed.
- Linked to parliamentary sources where available.
- Explained as a product summary rather than an official record.
Discouraged Claims
These claims need review because they can overstate what the product can prove:
- Neutral in an unqualified sense.
- Objective without explaining the source and review limits.
- Comprehensive when the product is curating or summarising.
- Verified unless a specific review standard has been completed.
- Safe without naming the safety boundary.
Banned Absolute Claims
Parliament People should not describe AI-assisted content as neutral, objective, error-free, official, or unbiased in absolute terms.
Safer Alternatives
| Avoid | Prefer | | --- | --- | | Official summary | Parliament People summary of parliamentary source material | | Neutral explanation | Source-grounded explanation | | Objective AI | AI-assisted text with visible provenance | | Error-free | Reviewable and corrected when needed | | Unbiased | Designed to preserve attribution and context |
Review Examples
Claims need review when they imply Parliament People has legal authority, perfect accuracy, total coverage, political neutrality, or independent verification that the evidence does not support.
Review Gate
Before publication, public copy that explains AI output should be checked against this standard. If a claim sounds stronger than the evidence or provenance supports, it should be rewritten.