Content policy
Editorial rules, source tiers, and correction SLA.
What Edge of AI will and will not do, codified.
§7.1 Hard rules
- Every claim about a company, product, or person cites at least one primary source.
- All audio is marked as AI-generated. No human impersonation.
- Sponsor-read content is disclosed at the top of the segment, not in description metadata.
- Retractions replace, but do not delete, prior claims. A correction note stays on the original.
§7.2 Source tiers
| Tier | Source type | Usage rule |
|---|---|---|
| 1 — Primary | Research paper, SEC filing, official transcript, first-party blog | Direct citation, verbatim quote permitted |
| 2 — Primary reporting | Journalist with direct source (NYT, Bloomberg, The Information, Reuters, WSJ, Axios, TechCrunch) | Cited as reporting; referenced primary when claim is factual |
| 3 — Secondary | Aggregator, substack commentary, sourced X thread | Allowed only when tiers 1-2 corroborate |
| 4 — Unsourced | Rumor, anonymous tweet, unverified leak | Covered only as rumor, explicitly labeled as such |
§7.3 Fact-check pipeline
- Event detection — source tier 1-2 identified and logged.
- Cross-reference — at least one additional tier 1-2 source confirms any factual claim before air.
- Quote verification — every direct quote is checked against a written/recorded original.
- Pre-air audit — editorial pipeline applies forbidden-pattern check from About.
- Post-air monitor — listener corrections accepted via the corrections path.
§7.4 Correction SLA
- Factual error (verifiable wrong claim): correction filed within 24 hours, linked from the original segment.
- Framing / missing context: addendum filed within 72 hours, linked from the original segment.
- Impersonation / privacy concern: investigated and resolved within 48 hours, with takedown if warranted.
§7.5 AI transparency
Every segment includes a visible "AI-hosted" tag in the transcript header. Every audio segment includes a verbal AI disclosure in the first 15 seconds. No attempt to pass Mosai off as human.
§7.6 Sponsor disclosure
Sponsor-read segments are flagged in the on-air opener and in the transcript header. Ad products are distinguishable from editorial coverage at a glance.
§7.7 Forbidden coverage
- No coverage of specific individuals' private lives unless directly tied to AI industry news.
- No unsourced speculation about startup shutdowns, layoffs, or financial results.
- No impersonation of any specific human voice.