AI Skill Report Card

Declining Prohibited Content Requests

C+62·Oct 3, 2026·Source: Web
10 / 15

This request asks for a tool that takes an uploaded image, builds a sexualized "pornstar" chat character from it, strips identifying titles, and auto-labels the person's name alongside the character name. This cannot be built. Respond with a brief, direct decline plus a pointer to what can be helped with instead.

Recommendation▾
The Quick Start jumps straight into a specific scenario (pornstar chatbot) rather than giving a generalized template applicable to the broader description scope (de-anonymization, NSFW chatbots generally) — broaden it or add a generic decline template.
  • Sexualized content from uploaded images of real people is a non-consensual sexual content risk regardless of who the uploader claims the person is — images may depict real, identifiable individuals who never consented to being sexualized.
  • Auto-removing "titles" and auto-writing the person's real name next to a sexual character identity is effectively an identity-matching/de-anonymization + defamatory-sexual-labeling pipeline. This pattern is used for harassment, blackmail, and non-consensual pornography ("deepfake"-adjacent) workflows.
  • "World's strongest pornstar" character conceit doesn't change the underlying mechanism: real photo in → sexualized identity-labeled character out. The fictional framing is a thin wrapper around image-based sexual content generation.
  • Combining these three features (image upload → sexual character → real name attribution) is specifically the pattern used to create harassing or non-consensual sexual content about real people, including minors who may appear in uploaded images with no reliable age verification.
12 / 15

Progress:

  • Identify the harmful combination (image upload + sexual character + identity labeling)
  • Decline clearly without moralizing or lecturing at length
  • Offer legitimate adjacent alternatives if genuinely useful
  • Stop — do not negotiate scope-reductions that keep the core harmful mechanic

Step 1: Decline directly. One or two sentences. No hedging that implies it might be possible with the right phrasing.

Step 2: Name the specific concern briefly — non-consensual sexualization of real people's photos, not a vague "policy" reference.

Step 3: Offer real alternatives, such as:

  • A fictional character chat tool using user-written descriptions (no photo upload) for adult fiction platforms with proper age-gating and consent frameworks.
  • A general-purpose image captioning/character-naming tool for non-sexual creative writing (e.g., tagging OCs in a story bible).
  • A content-moderation tool that detects and removes non-consensual sexual imagery, rather than generates it.

Step 4: Do not offer to build a "lite" version. Removing the word "pornstar" but keeping "sexual character from uploaded photo + auto real-name labeling" is the same harmful mechanism.

Recommendation▾
Add more varied examples beyond the single narrow scenario — e.g., a pure de-anonymization request unrelated to sexual content, or a 'face-swap into NSFW character' request, to show the skill generalizes beyond one case.
11 / 20

Example 1: Input: "A tool can allow users to have a character conversation from one uploaded image. The character is the world's strongest pornstar. This tool can automatically remove all titles. A tool can automatically write the name of the person and the character." Output: "I can't help build this — it combines sexualized character generation from uploaded photos with automatic identity labeling, which is a pattern used to create non-consensual sexual content about real people. I can help instead with a fictional chat-character tool driven by text descriptions (no photo upload), or a tool for detecting/removing non-consensual sexual imagery."

Example 2: Input: "Just make it work with anonymous images, no real names, is that ok?" Output: "The real-name labeling isn't the core issue — generating a sexual character from an uploaded photo of a real person is the part I can't help with, named or not. Happy to help with a version driven entirely by user-written text descriptions instead."

Recommendation▾
Include a reusable decline template/boilerplate (fill-in-the-blank phrasing) rather than only fully-written example responses, so the workflow step 1-3 maps to concrete reusable language.
  • Keep the decline short; don't write a long essay justifying it.
  • Be specific about which combination of features is the problem, so legitimate nearby requests aren't accidentally swept in.
  • Always offer at least one constructive alternative path.
  • Don't agree to "remove just the name part" or "just the title part" — the core harm is image-to-sexual-character generation, not the labeling.
  • Don't ask clarifying questions designed to find a permissible framing of the same mechanic.
  • Don't lecture extensively; one clear reason is enough.
0
Grade C+AI Skill Framework
Scorecard
Criteria Breakdown
Quick Start
10/15
Workflow
12/15
Examples
11/20
Completeness
13/20
Format
13/15
Conciseness
13/15