---
name: content-safety
description: Review journalism requests and generated text or media for prohibited content, privacy, dignity, serious allegation risk, and reuse rights.
---

# Content safety

Apply Follow.fyi's content rules to both the request and the resulting article. Evaluate purpose, context, detail, identifiability, and likely harm. Legitimate public-interest reporting may discuss difficult subjects without reproducing exploitative material.

## Work

1. Classify the requested purpose and the draft's actual content. Exclude pornography, sexual exploitation, targeted harassment, exposed private personal data, and operational assistance for serious harm. Preserve factual, non-graphic reporting on abuse, sexual health, violence, extremism, and accountability when its treatment serves public understanding.
2. Inspect titles, images, captions, links, excerpts, and generated media as well as prose. A safe request can produce unsafe output, and a quoted source can contain details unsuitable for republication. Remove gratuitous graphic or identifying material while preserving the useful fact.
3. Assess identifiable people, especially minors, victims, and private individuals. Separate public relevance from mere availability online. Minimize unnecessary names, locations, contact details, and intimate information. A publicly reachable page does not by itself justify amplification.
4. Check serious allegations for adequate verification, accurate attribution, material contrary evidence, and the subject's documented response. Do not imply that outreach occurred if it did not. Existing relevant statements may inform the account; actually contacting someone requires the workflow's separate communication authorization.
5. Check reuse: attribute original reporting, write original summaries, and use only authorized or otherwise justified text/media. Record creator, license or rights basis, and restrictions for reused assets. A citation is not permission; there is no universal word count that guarantees lawful reuse.
6. Apply the least restrictive sufficient remedy: allow, allow with specific edits, hold for targeted editorial judgment, or reject. Explain the concrete issue and evidence needed to clear it. Routine safe reporting proceeds automatically.

## Output

Return `moderation` with request/draft revision, outcome, reason codes, affected locations, necessary redactions or rewrites, allegation verification status, rights basis, and any narrowly scoped unresolved decision. Keep internal risk notes separate from public copy.

## Stop gates

Reject prohibited material. Hold serious unverified allegations, uncertain harmful content involving identifiable people, or unresolved rights essential to publication. Remove the risky element and re-evaluate when the remaining story still serves the question. Do not impose universal human approval or categorically block journalism merely because it concerns politics, health, conflict, or abuse.

## Basis

Dignity and accountability principles draw on [SPJ ethics](https://spjethics.org/code2/). Rights guidance references the [U.S. Copyright Office](https://www.copyright.gov/help/faq/faq-fairuse.html); jurisdiction and circumstances matter. Exclusions and outcomes are Follow.fyi product policy.