Content review processes and procedures
Pre-Screening and Post-Screening Policy
Last modified July 27, 2026
This Pre-Screening and Post-Screening Policy describes the timing of HoneyPot's ("HoneyPot's," "our," "we," or "us") content review processes: what we check before content is generated or published, and what we check after.
1. Pre-Screening
Pre-screening happens before content is generated, published to a public surface, or shown to other users.
Every user prompt is pre-screened. When you submit a prompt to a Honey:
- The prompt is passed through our classifier stack (see Content Moderation Policy) before being routed to a generation model
- Prompts that the classifier scores as high-confidence violations of the Prohibited Content Policy are blocked; you'll see a clear in-product message explaining the block
- Intermediate-confidence prompts may be allowed through with stricter generation-time safety constraints
Every new Honey is pre-screened. When you create a Honey, the name, appearance descriptors, personality prompt, and any uploaded gallery items are run through the same classifier stack before the Honey is saved. Configurations that violate the Prohibited Content Policy are blocked at save time.
Every Community-publish action is pre-screened. Before a user-created Honey appears in the public catalog, it goes through an additional review pass (automated + human in ambiguous cases). Until that review completes the Honey stays in your private library only.
Every uploaded media file is pre-screened. Avatar, cover, and gallery uploads are hashed against known illegal-content databases (NCMEC PhotoDNA for CSAM, GIFCT for terrorist content) before they're accepted. Matches are blocked and, for CSAM, reported to NCMEC.
2. Post-Screening
Post-screening happens after content has been generated or published, as an additional safety net.
Every AI generation is post-screened. When a Honey produces text, image, or video output:
- The output is re-scanned by classifiers tuned to the generation modality
- Images are hashed against the same illegal-content databases used in pre-screening
- High-confidence violations are quarantined and the conversation is reviewed by a human moderator
Public surfaces are sampled for ongoing review. The community Honey catalog, public profiles, and any other publicly browsable surface are continuously sampled by automated systems and reviewed by human moderators. Findings can lead to retroactive removal (see Content Removal Policy).
Reports trigger post-screening review. Every user, third-party, or law-enforcement report kicks off a focused post-screening review of the reported content and, where appropriate, the reporting user's broader content and account.
3. Reviewers
Pre- and post-screening reviewers operate under written guidelines tied to the Prohibited Content Policy and the Content Moderation Policy. Reviewers:
- Are background-checked
- Sign confidentiality agreements
- Have access to context (Honey configuration, conversation history) sufficient to make accurate calls
- Are subject to QA sampling and regular policy retraining
- Have access to mental-health resources given the nature of the role
4. Decision Times
| Surface | Pre-screening target | Post-screening target |
|---|---|---|
| User prompt | Real-time (< 500 ms) | Real-time (< 500 ms after generation) |
| New Honey creation | Real-time (< 5 seconds) | Sampled within 24 hours |
| Community publish | Within 24 hours | Sampled continuously |
| Media upload | Real-time hash check | Sampled within 24 hours |
| Reported content | n/a | Highest priority: within 24 hours for CSAM / non-consensual / threat reports; within 5 business days for other categories |
5. False Positives and Appeals
No automated system is perfect. If you believe your content was incorrectly blocked or removed, follow the appeals process in the Content Removal Policy. Confirmed false positives are fed back into our classifier training so the system improves over time.
6. Continuous Improvement
We iterate on classifiers, hash databases, and reviewer guidelines on an ongoing basis as new threats and content patterns emerge. The Content Moderation Policy describes the broader moderation program of which pre- and post-screening are a part.