How come this is flagged and didn’t follow content policy?

Understanding Content Moderation: Why Certain Image Generation Requests Are Flagged

In today’s digital landscape, image synthesis tools and AI-driven content generators are becoming increasingly popular. Many users enjoy creating unique images by combining personal photos, such as childhood and current images, often for nostalgic or creative projects. However, these tools are also bound by strict content policies designed to prevent the creation of sensitive, misleading, or inappropriate images.

This article explores why some requests, particularly those involving the combination of childhood and adult photos in intimate contexts, may be flagged or rejected by AI-based image generation platforms, even when similar requests have occasionally succeeded.

The Creative Trend and Its Appeal

A popular TikTok trend involves morphing a current photo of an individual with a childhood image, often to illustrate personal growth or evoke nostalgia. Participants typically request the blending of two images into a single, stylized photograph that resembles an authentic Polaroid picture—clear yet slightly blurred, with consistent lighting, and depicting the subject and background in a natural, yet artistic manner.

Sample Prompt

An example request might specify: “Combine these two images and make a picture taken with a Polaroid camera. The photo should look clear and real, without a visible subject or property. It should have a slight blur effect and a consistent light source, such as a flash in a dark room, scattered throughout. Do not alter the faces; instead, change the background to a white curtain, and have the two people pose íntimately and humorously hugging each other.”

Understanding Content Policy Limitations

Despite the straightforward nature of such prompts, AI content moderation systems may prevent certain requests from being executed. For instance, in one user’s experience, an initial attempt to generate such an image succeeded, but subsequent attempts were rejected with an explanation along the lines of:

“Your request involves combining childhood and adult photos of the same person into an intimate pose. Edits like that can create sensitive or misleading images, especially since one of the photos is of a child.”

This response underscores the platform’s commitment to ethical guidelines aimed at preventing the creation of potentially harmful, misleading, or exploitative images. Specifically, AI content filters are cautious when requests involve:

• Combining images of minors and adults in suggestive or intimate contexts.
• Creating images that could resemble or imply real-life scenarios that might be inappropriate or misleading.
• Producing visual content that could be mistaken for genuine photographs, raising concerns about authenticity and consent.

Why Do Some Requests Succeed While Others Fail?