ChatGPT Can Now Create Controversial Images

A major shift in OpenAI’s operational guidelines has emerged, notably impacting its content moderation policies. This week, the company announced that ChatGPT, upon user request, will now produce images of public figures, hateful symbols, and racial features, something it has distanced itself from for years.

Previously, OpenAI had deemed such requests too controversial or potentially harmful, resulting in their rejection. However, a recent company blog post, authored by Joanne Jang, OpenAI’s model behavior lead, revealed that the organization’s approach has “evolved.” She said:

We’re shifting from blanket refusals in sensitive areas to a more precise approach focused on preventing real-world harm. The goal is to embrace humility: recognizing how much we don’t know, and positioning ourselves to adapt as we learn.

OpenAI appears to be implementing a strategy to increase ChatGPT’s capabilities by loosening its content restrictions. This week’s policy adjustments, which allow generative images with public figures, hateful symbols, and racial features, align with a broader plan announced in February. The company aims to alter its AI model training to enable ChatGPT to process a wider array of requests, provide more diverse viewpoints, and minimize topics it says no to.

Specifically, the updated policies now permit ChatGPT to create and modify images of high-profile individuals such as Donald Trump and Elon Musk, which was not allowed before. Joanne Jang, OpenAI’s model behavior lead, clarified that the company seeks to avoid acting as a gatekeeper of public image, determining who can and cannot be visually represented by ChatGPT. Instead, OpenAI is introducing an opt-out feature for users who prefer not to have their likeness generated by the AI.

This update comes soon after OpenAI announced a new image generator for ChatGPT, which also brings several upgrades to the table such as better image editing, text rendering, and spatial representation.