AI disclosure
How Clerisi may use AI as an editorial assistant, what is not automated, and where human review remains mandatory.
1. AI is an assistant, not the authority
Where configured, Clerisi may use AI to assist parts of editorial preparation. AI does not determine truth, decide event outcomes, make final editorial judgments, or replace accountable human review.
Public promise: AI can help structure the record. It should not quietly become the record.
2. Where AI may assist
Transcripts
Draft transcript cleanup, speaker-label suggestions, and timestamp organisation where a transcript workflow is in use.
Summaries
Initial summaries of arguments, objections, claims, evidence gaps, and session outcomes.
Metadata
Topic tags, archive descriptions, related content suggestions, and search metadata.
Clips
Candidate highlight moments, quote extraction, and short-form editorial prompts.
3. What AI should not do alone
- Issue final Idea Court assessments without accountable human review.
- Label a claim as true, false, settled, or disproven without editorial evidence review.
- Invent citations, evidence, quotes, or speaker positions.
- Silently rewrite speaker arguments in a way that changes meaning.
- Replace correction, appeal, or editorial accountability processes.
4. Human review requirements
Clerisi outputs that influence public interpretation should go through human editorial review. This includes event summaries, claim-status labels, evidence-gap analysis, article derivatives, resource briefs, and archive descriptions.
5. Disclosure labels
Where AI assistance materially shaped a public output, Clerisi should include a disclosure label such as “AI-assisted transcript reviewed by Clerisi” or “AI-assisted metadata.”
6. Errors and correction
If AI-assisted material introduces a factual, attribution, or interpretation error, Clerisi should correct the public record and identify the affected output.
7. Future intelligence layer
As Clerisi develops its public idea-testing intelligence system, AI may support argument maps, evidence-gap analysis, objection patterns, and idea-maturity interpretation. These are editorial intelligence products, not neutral machine truth.