OpenAI Reverses Ban on Face Editing in DALL-E 2
OpenAI now allows DALL-E 2 users to upload and edit photos of people, citing improved safety systems to prevent deepfakes and abuse. The company maintains strict consent rules as it aims to avoid the controversies plaguing rival image generators.
OpenAI now permits DALL-E 2 users to upload and edit images of people, reversing a previous ban on photorealistic faces. The company states that upgraded safety systems minimize the risks of deepfakes and prevent the creation of harmful sexual, political, or violent content. Customers, ranging from reconstructive surgeons to filmmakers, express excitement about using the tool to visualize medical results and accelerate creative workflows.
Despite this policy shift, OpenAI does not allow unrestricted face manipulation. The terms of service strictly prohibit users from uploading pictures of individuals without their explicit consent or using images without proper rights. However, the actual enforcement of these rules remains uncertain, making this rollout a significant test of the company's filtering technology, which some users previously criticize as overly aggressive or inaccurate.
This cautious approach differentiates OpenAI from competitors like Stability AI, whose open-source Stable Diffusion model lacks usage restrictions. Unfettered access to rival tools quickly leads to the creation of nonconsensual celebrity deepfakes, prompting OpenAI to maintain its buttoned-up, brand-friendly image. By keeping safeguards in place, OpenAI hopes to provide creative freedom while avoiding the severe controversies that come with unregulated AI image synthesis.